pyvale 2025.4.0__py3-none-any.whl

pyvale/core/cython/rastercyth.py

@@ -0,0 +1,636 @@
+"""
+================================================================================
+pyvale: the python validation engine
+License: MIT
+Copyright (C) 2025 The Computer Aided Validation Team
+================================================================================
+"""
+import numpy as np
+import cython
+from cython.parallel import prange, parallel, threadid
+from cython.cimports.libc.math import floor, ceil
+
+from pyvale.core.rendermesh import RenderMeshData
+from pyvale.core.cameradata import CameraData
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def range_len_double(start: cython.double,
+                     stop: cython.double,
+                     step: cython.double) -> cython.size_t:
+    return int(ceil((stop - start) / step))
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def vec_range_int(start: cython.int,
+                  stop: cython.int,
+                  step: cython.int,
+                  vec_buffer: cython.long[:]) -> cython.long[:]:
+
+    num_vals: cython.size_t = int(ceil((stop - start) / step))
+
+    vec_buffer[0] = start
+    ii: cython.size_t
+    for ii in range(1,num_vals):
+        vec_buffer[ii] = vec_buffer[ii-1] + step
+
+    return vec_buffer[0:num_vals]
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def vec_max_double(vals: cython.double[:]) -> cython.double:
+
+    num_vals: cython.size_t = vals.shape[0]
+
+    ii: cython.size_t = 0
+    max_val: cython.double = vals[ii]
+
+    for ii in range(1,num_vals):
+        if vals[ii] > max_val:
+            max_val = vals[ii]
+
+    return max_val
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def vec_min_double(vals: cython.double[:]) -> cython.double:
+
+    num_vals: cython.size_t = vals.shape[0]
+
+    ii: cython.size_t = 0
+    min_val: cython.double = vals[ii]
+
+    for ii in range(1,num_vals):
+        if vals[ii] < min_val:
+            min_val = vals[ii]
+
+    return min_val
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def vec_dot_double(vec0: cython.double[:], vec1: cython.double[:]
+                   ) -> cython.double:
+    vec0_len: cython.size_t = vec0.shape[0]
+    vec1_len: cython.size_t = vec1.shape[0]
+    if vec0_len != vec1_len:
+        return 0.0
+
+    ii: cython.size_t = 0
+    dot: cython.double = 0.0
+    for ii in range(vec0_len):
+        dot += vec0[ii]*vec1[ii]
+
+    return dot
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def bound_index_min(min_val: cython.double) -> cython.int:
+    min_ind: cython.int = int(floor(min_val))
+    if min_ind < 0:
+        min_ind = 0
+    return min_ind
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def bound_index_max(max_val: cython.double,
+                    num_pixels: cython.int) -> cython.int:
+    max_ind: cython.int = int(ceil(max_val))
+    if max_ind > (num_pixels-1):
+        max_ind = (num_pixels-1)
+    return max_ind
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.inline
+@cython.exceptval(check=False)
+def mult_mat44_by_vec3(mat44: cython.double[:,:], vec3_in: cython.double[:],
+                       vec3_out: cython.double[:]) -> cython.double[:]:
+
+    vec3_out[0] = (mat44[0,0]*vec3_in[0]
+                      + mat44[0,1]*vec3_in[1]
+                      + mat44[0,2]*vec3_in[2]
+                      + mat44[0,3])
+    vec3_out[1] = (mat44[1,0]*vec3_in[0]
+                      + mat44[1,1]*vec3_in[1]
+                      + mat44[1,2]*vec3_in[2]
+                      + mat44[1,3])
+    vec3_out[2] = (mat44[2,0]*vec3_in[0]
+                      + mat44[2,1]*vec3_in[1]
+                      + mat44[2,2]*vec3_in[2]
+                      + mat44[2,3])
+    vec3_out[3] = (mat44[3,0]*vec3_in[0]
+                      + mat44[3,1]*vec3_in[1]
+                      + mat44[3,2]*vec3_in[2]
+                      + mat44[3,3])
+
+    return vec3_out
+
+
+@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.exceptval(check=False)
+def world_to_raster_coords(coords_world: cython.double[:],
+                           world_to_cam_mat: cython.double[:,:],
+                           image_dist: cython.double,
+                           image_dims: cython.double[:],
+                           num_pixels: cython.int[:],
+                           coords_raster: cython.double[:]
+                           ) -> cython.double[:]:
+    xx: cython.size_t = 0
+    yy: cython.size_t = 1
+    zz: cython.size_t = 2
+    ww: cython.size_t = 3
+
+    coords_raster = mult_mat44_by_vec3(world_to_cam_mat,
+                                       coords_world,
+                                       coords_raster)
+
+    coords_raster[xx] = coords_raster[xx] / coords_raster[ww]
+    coords_raster[yy] = coords_raster[yy] / coords_raster[ww]
+    coords_raster[zz] = coords_raster[zz] / coords_raster[ww]
+
+    coords_raster[xx] = (image_dist * coords_raster[xx]
+                        / -coords_raster[zz])
+    coords_raster[yy] = (image_dist * coords_raster[yy]
+                        / -coords_raster[zz])
+
+    coords_raster[xx] = 2*coords_raster[xx] / image_dims[xx]
+    coords_raster[yy] = 2*coords_raster[yy] / image_dims[yy]
+
+    coords_raster[xx] = (coords_raster[xx] + 1)/2 * num_pixels[xx]
+    coords_raster[yy] = (1-coords_raster[yy])/2 * num_pixels[yy]
+    coords_raster[zz] = -coords_raster[zz]
+
+    return coords_raster
+
+@cython.cfunc
+@cython.nogil
+@cython.boundscheck(False)
+@cython.wraparound(False)
+@cython.inline
+@cython.exceptval(check=False)
+def edge_function(vert_0: cython.double[:],
+                  vert_1: cython.double[:],
+                  vert_2: cython.double[:]) -> cython.double:
+    edge_fun: cython.double = (
+        (vert_2[0] - vert_0[0]) * (vert_1[1] - vert_0[1])
+        - (vert_2[1] - vert_0[1]) * (vert_1[0] - vert_0[0]))
+    return edge_fun
+
+
+@cython.cfunc
+@cython.nogil
+@cython.boundscheck(False)
+@cython.wraparound(False)
+@cython.inline
+@cython.exceptval(check=False)
+def edge_function_pt(vert_0: cython.double[:],
+                     vert_1: cython.double[:],
+                     vert_2_x: cython.double,
+                     vert_2_y: cython.double) -> cython.double:
+    edge_fun: cython.double = (
+        (vert_2_x - vert_0[0]) * (vert_1[1] - vert_0[1])
+        - (vert_2_y - vert_0[1]) * (vert_1[0] - vert_0[0]))
+    return edge_fun
+
+
+@cython.ccall
+@cython.boundscheck(False)
+@cython.wraparound(False)
+@cython.cdivision(True)
+def average_image(image_subpx: cython.double[:,:],
+                  sub_samp: cython.int,
+                  ) -> cython.double[:,:]:
+
+    if sub_samp <= 1:
+        return np.asarray(image_subpx[:,:])
+
+    px_num_y: cython.size_t = int(ceil(image_subpx.shape[0]/sub_samp))
+    px_num_x: cython.size_t = int(ceil(image_subpx.shape[1]/sub_samp))
+
+    image_buff_avg_np = np.full((px_num_y,px_num_x),0.0,dtype=np.float64)
+    image_buff_avg: cython.double[:,:] = image_buff_avg_np
+
+    num_subpx_y: cython.size_t = image_subpx.shape[0]
+    num_subpx_x: cython.size_t = image_subpx.shape[1]
+    subpx_per_px: cython.double = float(sub_samp*sub_samp)
+    ss_size: cython.size_t = sub_samp
+
+    num_px_y: cython.size_t = int(num_subpx_y/sub_samp)
+    num_px_x: cython.size_t = int(num_subpx_x/sub_samp)
+
+    px_sum: cython.double = 0.0
+
+    ix: cython.size_t = 0
+    iy: cython.size_t = 0
+    sx: cython.size_t = 0
+    sy: cython.size_t = 0
+
+    for iy in range(num_px_y):
+        for ix in range(num_px_x):
+            px_sum = 0.0
+            for sy in range(ss_size):
+                for sx in range(ss_size):
+                    px_sum += image_subpx[ss_size*iy+sy,ss_size*ix+sx]
+
+            image_buff_avg[iy,ix] = px_sum / subpx_per_px
+
+    return image_buff_avg
+
+
+@cython.nogil
+@cython.cfunc
+@cython.boundscheck(False)
+@cython.wraparound(False)
+@cython.cdivision(True)
+@cython.exceptval(check=False)
+def _average_image(image_buff_subpx_in: cython.double[:,:],
+                   sub_samp: cython.int,
+                   image_buff_avg_out: cython.double[:,:]
+                   ) -> cython.int:
+
+    num_subpx_y: cython.size_t = image_buff_subpx_in.shape[0]
+    num_subpx_x: cython.size_t = image_buff_subpx_in.shape[1]
+    subpx_per_px: cython.double = float(sub_samp*sub_samp)
+    ss_size: cython.size_t = sub_samp
+
+    num_px_y: cython.size_t = int(num_subpx_y/sub_samp)
+    num_px_x: cython.size_t = int(num_subpx_x/sub_samp)
+
+    px_sum: cython.double = 0.0
+
+    ix: cython.size_t = 0
+    iy: cython.size_t = 0
+    sx: cython.size_t = 0
+    sy: cython.size_t = 0
+
+    for iy in range(num_px_y):
+        for ix in range(num_px_x):
+            px_sum = 0.0
+            for sy in range(ss_size):
+                for sx in range(ss_size):
+                    px_sum += image_buff_subpx_in[ss_size*iy+sy,ss_size*ix+sx]
+
+            image_buff_avg_out[iy,ix] = px_sum / subpx_per_px
+
+    return 0
+
+
+#///////////////////////////////////////////////////////////////////////////////
+@cython.ccall # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+def raster_frame(coords: cython.double[:,:],
+                 connect: cython.size_t[:,:],
+                 fields_to_render: cython.double[:,:],
+                 cam_data: CameraData,
+                 ) -> tuple[np.ndarray,np.ndarray,int]:
+
+
+    world_to_cam_mat: cython.double[:,:] = cam_data.world_to_cam_mat
+    pixels_num: cython.int[:] = cam_data.pixels_num
+    image_dims: cython.double[:] = cam_data.image_dims
+    image_dist: cython.double = cam_data.image_dist
+    sub_samp: cython.int = cam_data.sub_samp
+
+    nodes_per_elem: cython.size_t = connect.shape[1]
+    fields_num: cython.size_t = fields_to_render.shape[1]
+    sub_pix_x: cython.int = pixels_num[0]*sub_samp
+    sub_pix_y: cython.int = pixels_num[1]*sub_samp
+
+    #---------------------------------------------------------------------------
+    # Final image buffer memory allocation
+    image_buff_avg_np = np.full((pixels_num[1],pixels_num[0],fields_num),0.0,dtype=np.float64)
+    image_buff_avg: cython.double[:,:,:] = image_buff_avg_np
+
+    depth_buff_avg_np = np.full((pixels_num[1],pixels_num[0]),0.0,dtype=np.float64)
+    depth_buff_avg: cython.double[:,:] = depth_buff_avg_np
+
+    #---------------------------------------------------------------------------
+    # Per-thread scratch memory allocations
+    depth_buffer_np = np.full((sub_pix_y,sub_pix_x),1.0e6,dtype=np.float64)
+    depth_buff_subpx: cython.double[:,:] = depth_buffer_np
+
+    image_buffer_np = np.full((sub_pix_y,sub_pix_x,fields_num),0.0,dtype=np.float64)
+    image_buff_subpx: cython.double[:,:,:] = image_buffer_np
+
+    # shape=(nodes_per_elem, coord[X,Y,Z,W])
+    nodes_raster_np = np.empty((nodes_per_elem,4),dtype=np.float64)
+    nodes_raster_buff: cython.double[:,:] = nodes_raster_np
+
+    field_raster_np = np.empty((nodes_per_elem,),dtype=np.float64)
+    field_raster_buff: cython.double[:] = field_raster_np
+
+    px_coord_np = np.zeros((nodes_per_elem,),np.float64)
+    px_coord_buff: cython.double[:] = px_coord_np
+
+    weights_np = np.zeros((nodes_per_elem,),np.float64)
+    weights_buff: cython.double[:] = weights_np
+    #---------------------------------------------------------------------------
+
+    elems_in_image: cython.size_t = _raster_frame(coords[:,:],
+                                                connect[:,:],
+                                                fields_to_render[:,:],
+                                                world_to_cam_mat[:,:],
+                                                pixels_num[:],
+                                                image_dims[:],
+                                                image_dist,
+                                                sub_samp,
+                                                image_buff_avg[:,:,:],
+                                                depth_buff_avg[:,:],
+                                                image_buff_subpx[:,:,:],
+                                                depth_buff_subpx[:,:],
+                                                nodes_raster_buff[:,:],
+                                                field_raster_buff[:],
+                                                px_coord_buff[:],
+                                                weights_buff[:])
+
+    return (image_buff_avg_np,depth_buff_avg_np,elems_in_image)
+
+
+#///////////////////////////////////////////////////////////////////////////////
+#@cython.nogil
+@cython.cfunc # python+C or cython.cfunc for C only
+@cython.boundscheck(False) # Turn off array bounds checking
+@cython.wraparound(False)  # Turn off negative indexing
+@cython.cdivision(True)    # Turn off divide by zero check
+@cython.exceptval(check=False) # Turn off exceptions
+def _raster_frame(coords: cython.double[:,:],
+                 connect: cython.size_t[:,:],
+                 fields_to_render: cython.double[:,:],
+                 world_to_cam_mat: cython.double[:,:],
+                 num_pixels: cython.int[:],
+                 image_dims: cython.double[:],
+                 image_dist: cython.double,
+                 sub_samp: cython.int,
+                 # From here these are memory buffers that will be written into
+                 image_buff_avg: cython.double[:,:,:],
+                 depth_buff_avg: cython.double[:,:],
+                 image_buff_subpx: cython.double[:,:,:],
+                 depth_buff_subpx: cython.double[:,:],
+                 nodes_raster_buff: cython.double[:,:],
+                 field_raster_buff: cython.double[:],
+                 px_coord_buff: cython.double[:],
+                 weights_buff: cython.double[:],
+                 ) -> cython.size_t:
+    """Rasters a single frame and all associated fields into the image and depth
+    buffer provided as inputs to the function. This is a pure cython function
+    with the GIL released for parallelisation. All fields (textures) are
+    rendered in a sub-loop so that the depth buffer and inside/outside test is
+    only performed once for all fields to be rendered.
+
+    Parameters
+    ----------
+    coords : cython.double[:,:]
+        Input. shape=(num_nodes,coords[x,y,z,w])
+    connect : cython.size_t[:,:]
+        Input. shape=(num_elems,nodes_per_elem)
+    fields_to_render : cython.double[:,:]
+        Input. shape=(num_nodes,num_fields)
+    world_to_cam_mat : cython.double[:,:]
+        Input. Homogeneous coordinate transformation matrix from world to camera
+         coordinates. shape=(4,4).
+    num_pixels : cython.int[:]
+        Input. shape=(2 [num_px_x,num_px_y],)
+    image_dims : cython.double[:]
+        Input. shape=(2 [fov_size_x,fov_size_y],)
+    image_dist : cython.double
+        Input.
+    sub_samp : cython.int
+        Number of subsamples per pixel for anti-aliasing.
+    image_buff_avg : cython.double[:,:,:]
+        Output buffer. shape=(num_px_y,num_px_x,num_fields)
+    depth_buff_avg : cython.double[:,:]
+        Output buffer. shape=(num_px_y,num_px_x)
+    image_buff_subpx : cython.double[:,:,:]
+        Processing buffer (output). shape=(num_subpx_y,num_subpx_x,num_fields)
+    depth_buff_subpx : cython.double[:,:]
+        Processing buffer (output). shape=(num_subpx_y,num_subpx_x)
+    nodes_raster_buff : cython.double[:,:]
+        Processing buffer (output). shape=(nodes_per_elem, 4 coord[x,y,z,w])
+    field_raster_buff : cython.double[:]
+        Processing buffer (output). shape=(nodes_per_elem,)
+    px_coord_buff : cython.double[:]
+        Processing buffer (output). shape=(nodes_per_elem,)
+    weights_buff : cython.double[:]
+        Processing buffer (output). shape=(nodes_per_elem,)
+
+    Returns
+    -------
+    cython.size_t
+        Number of rendered elements after backface culling and cropping.
+    """
+
+    xx: cython.size_t = 0
+    yy: cython.size_t = 1
+    zz: cython.size_t = 2
+
+    elem_count: cython.size_t = connect.shape[0]
+    nodes_per_elem: cython.size_t = connect.shape[1]
+    fields_num: cython.size_t = fields_to_render.shape[1]
+
+    # tolerance for floating point zero dot product
+    tol: cython.double = 1e-12
+
+    #elem_count: cython.size_t = 1
+    elems_in_image: cython.size_t = 0
+
+    ee: cython.size_t = 0
+    nn: cython.size_t = 0
+    ii: cython.size_t = 0
+    jj: cython.size_t = 0
+    ww: cython.size_t = 0
+    ff: cython.size_t = 0
+
+    for ee in range(elem_count):
+
+        for nn in range(nodes_per_elem):
+            # shape=(nodes_per_elem, coord[X,Y,Z,W])
+            nodes_raster_buff[nn,:] = world_to_raster_coords(coords[connect[ee,nn],:],
+                                                        world_to_cam_mat,
+                                                        image_dist,
+                                                        image_dims,
+                                                        num_pixels,
+                                                        nodes_raster_buff[nn,:])
+
+
+        elem_area: cython.double = edge_function(nodes_raster_buff[0,:],
+                                                 nodes_raster_buff[1,:],
+                                                 nodes_raster_buff[2,:])
+
+        if elem_area < -tol: # Backface culling
+            continue
+
+        print(f"{nodes_raster_buff[0,0]},{nodes_raster_buff[0,1]},{nodes_raster_buff[0,2]}")
+        print(f"{nodes_raster_buff[1,0]},{nodes_raster_buff[1,1]},{nodes_raster_buff[1,2]}")
+        print(f"{nodes_raster_buff[2,0]},{nodes_raster_buff[2,1]},{nodes_raster_buff[2,2]}")
+        print(f"{ee} ELEM AREA : {elem_area}")
+        print()
+
+        x_min: cython.double = vec_min_double(nodes_raster_buff[:,xx])
+        x_max: cython.double = vec_max_double(nodes_raster_buff[:,xx])
+
+        if ((x_min > num_pixels[xx]-1) or (x_max < 0)): # x crop
+            continue
+
+        y_min: cython.double = vec_min_double(nodes_raster_buff[:,yy])
+        y_max: cython.double = vec_max_double(nodes_raster_buff[:,yy])
+
+        if ((y_min > num_pixels[yy]-1) or (y_max < 0)): # y crop
+            continue
+
+        elems_in_image += 1
+
+        xi_min: cython.size_t = bound_index_min(x_min)
+        xi_max: cython.size_t = bound_index_max(x_max,num_pixels[xx])
+        yi_min: cython.size_t = bound_index_min(y_min)
+        yi_max: cython.size_t = bound_index_max(y_max,num_pixels[yy])
+
+        for nn in range(nodes_per_elem):
+            nodes_raster_buff[nn,zz] = 1/nodes_raster_buff[nn,zz]
+
+        num_bound_x: cython.size_t = range_len_double(float(xi_min),
+                                                      float(xi_max),
+                                                      1.0/float(sub_samp))
+        num_bound_y: cython.size_t = range_len_double(float(yi_min),
+                                                      float(yi_max),
+                                                      1.0/float(sub_samp))
+
+        bound_coord_x: cython.double = float(xi_min) + 1.0/(2.0*float(sub_samp))
+        bound_coord_y: cython.double = float(yi_min) + 1.0/(2.0*float(sub_samp))
+        coord_step: cython.double = 1.0/float(sub_samp)
+        bound_ind_x: cython.size_t = sub_samp*xi_min
+        bound_ind_y: cython.size_t = sub_samp*yi_min
+
+
+        for jj in range(num_bound_y):
+
+            bound_coord_x = float(xi_min) + 1.0/(2.0*float(sub_samp))
+            bound_ind_x: cython.size_t = sub_samp*xi_min
+
+            for ii in range(num_bound_x):
+
+                px_coord_buff[xx] = bound_coord_x
+                px_coord_buff[yy] = bound_coord_y
+
+                # Check the edge functions for each edge one at a time, as soon
+                # as one is outside we don't need to do anymore work
+                weights_buff[0] = edge_function(nodes_raster_buff[1,:],
+                                                nodes_raster_buff[2,:],
+                                                px_coord_buff)
+                if (weights_buff[0] < -tol):
+                    bound_coord_x += coord_step
+                    bound_ind_x += 1
+                    continue
+
+                weights_buff[1] = edge_function(nodes_raster_buff[2,:],
+                                                nodes_raster_buff[0,:],
+                                                px_coord_buff)
+                if (weights_buff[1] < -tol):
+                    bound_coord_x += coord_step
+                    bound_ind_x += 1
+                    continue
+
+
+                weights_buff[2] = edge_function(nodes_raster_buff[0,:],
+                                                nodes_raster_buff[1,:],
+                                                px_coord_buff)
+                if (weights_buff[2] < -tol):
+                    bound_coord_x += coord_step
+                    bound_ind_x += 1
+                    continue
+
+
+                for ww in range(nodes_per_elem):
+                    weights_buff[ww] = weights_buff[ww] / elem_area
+
+                weight_dot_nodes: cython.double = vec_dot_double(
+                                                    weights_buff,
+                                                    nodes_raster_buff[:,zz])
+
+                # Check the depth buffer, if the element is behind move on
+                px_coord_z: cython.double = 1/weight_dot_nodes
+                if px_coord_z >= depth_buff_subpx[bound_ind_y,bound_ind_x]:
+                    continue
+
+                # We only need one depth buffer for all fields
+                depth_buff_subpx[bound_ind_y,bound_ind_x] = px_coord_z
+
+                for ff in range(fields_num):
+                    for nn in range(nodes_per_elem):
+
+                        field_raster_buff[nn] = (fields_to_render[connect[ee,nn],ff]
+                                                *nodes_raster_buff[nn,zz])
+
+                    px_field: cython.double = (vec_dot_double(field_raster_buff,
+                                                              weights_buff)
+                                                              *px_coord_z)
+
+                    image_buff_subpx[bound_ind_y,bound_ind_x,ff] = px_field
+
+                # end for(x) - increment the x coords
+                bound_coord_x += coord_step
+                bound_ind_x += 1
+
+            # end for(y) - increment the y coords
+            bound_coord_y += coord_step
+            bound_ind_y += 1
+
+    _average_image(depth_buff_subpx,sub_samp,depth_buff_avg)
+
+    for ff in range(fields_num):
+        _average_image(image_buff_subpx[:,:,ff],
+                        sub_samp,
+                        image_buff_avg[:,:,ff])
+
+    return elems_in_image