similie 0.2.6 → 0.3.1

data/README.rdoc

@@ -6,25 +6,34 @@ Similie is a simple DCT based image hashing interface that,
 * computes hamming distance between 2 fingerprints.
 
 == Example
-    
+
     require 'similie'
 
-    img1 = Similie.new("test/lena1.jpg") 
-    img2 = Similie.new("test/lena2.png") # lena1.jpg cropped and scaled
+    img1 = Similie.new("test/lena1.png")
+    img2 = Similie.new("test/lena2.png") # lena1.png cropped and scaled
     img3 = Similie.new("test/lena5.png") # a different image
 
-    img1.hash #=> unsigned 64bit int
+    img1.fingerprint #=> 64bit int
+
+    img1.distance(img2) #=> 2
+    img1.distance(img5) #=> 12
 
-    img1.distance(img2) #=> 1
-    img1.distance(img5) #=> 5
+    # class methods, if you want to deallocate image buffers immediately.
+    Similie.distance("test/lena1.png", "test/lena5.png") #=> 12
+    Similie.fingerprint("test/lena1.png")
 
-    Similie.distance("test/lena1.jpg", "test/lena5.jpg") #=> 5
+    # utility method that exposes hamming distance http://en.wikipedia.org/wiki/Hamming_weight
+    Similie.popcount(0x03 ^ 0x08) #=> 3
 
 == Dependencies
 
 * ruby 1.9.1+
 * opencv 2.1+  (libcv-dev and libhighgui-dev on debian systems)
 
+= See Also
+
+{pHash - The open source perceptual hash library}[http://www.phash.org/]
+
 = License
 
 {CC BY-SA 3.0}[http://creativecommons.org/licenses/by-sa/3.0/]

data/ext/extconf.rb

@@ -1,12 +1,25 @@
 require 'mkmf'
 
-$CFLAGS  = "-I/usr/include/opencv -Wall"
-$LDFLAGS = "-lhighgui -lcxcore"
+def inc_paths lib, defaults
+  path = %x{pkg-config #{lib} --cflags 2>/dev/null}.strip
+  path.size > 0 ? path : defaults.map {|name| "-I#{name}"}.join(' ')
+end
+
+def lib_paths lib, defaults
+  path = %x{pkg-config #{lib}  --libs-only-L 2>/dev/null}.strip
+  path.size > 0 ? path : defaults.map {|name| "-L#{name}"}.join(' ')
+end
+
+def lib_names lib, defaults
+  path = %x{pkg-config #{lib}  --libs-only-l 2>/dev/null}.strip
+  path.size > 0 ? path : defaults.map {|name| "-l#{name}"}.join(' ')
+end
 
-dir_config("libcv", ["/usr/local", "/opt/local", "/usr"])
+$CFLAGS  = inc_paths('opencv', %w(/usr/include/opencv)) + ' -Wall'
+$LDFLAGS = lib_names('opencv', %w(highgui cxcore))
 
 headers = [ 'stdio.h', 'stdlib.h', 'string.h', 'opencv/cxcore.h', 'opencv/highgui.h' ]
-lib_1   = [ 'cxcore', 'cvInitFont', headers ]
+lib_1   = [ 'cxcore',  'cvInitFont',    headers ]
 lib_2   = [ 'highgui', 'cvEncodeImage', headers ]
 
 if have_header('opencv/cxcore.h') && have_library(*lib_1) && have_library(*lib_2)

data/ext/similie.c

@@ -21,11 +21,8 @@
 #include <opencv/cxcore.h>
 #include <opencv/highgui.h>
 
-#define TO_S(v)                    rb_funcall(v, rb_intern("to_s"), 0)
-#define CSTRING(v)                 RSTRING_PTR(TO_S(v))
-#define ID_CONST_GET               rb_intern("const_get")
-#define CONST_GET(scope, constant) (rb_funcall(scope, ID_CONST_GET, 1, rb_str_new2(constant)))
-#define cvGetMonoPixel(img,y,x)    ((uchar *)(img->imageData + x*img->widthStep))[y*img->nChannels]
+#define TO_S(v)    rb_funcall(v, rb_intern("to_s"), 0)
+#define CSTRING(v) RSTRING_PTR(TO_S(v))
 
 #undef SIZET2NUM
 #ifdef HAVE_LONG_LONG
@@ -34,6 +31,11 @@
   #define SIZET2NUM(x) ULONG2NUM(x)
 #endif
 
+#define DCT_SIZE 32
+
+#if __GNUC__ >= 4 && __GNUC_MINOR__ >= 4
+#define popcount __builtin_popcountll
+#else
 // http://en.wikipedia.org/wiki/Hamming_weight
 
 const uint64_t m1  = 0x5555555555555555; //binary: 0101...
@@ -51,47 +53,39 @@ int popcount(uint64_t x) {
     x = (x + (x >> 4)) & m4;        //put count of each 8 bits into those 8 bits
     return (x * h01)>>56;           //returns left 8 bits of x + (x<<8) + (x<<16) + (x<<24) + ...
 }
+#endif
 
-#define DCT_SIZE 32
-
-uint64_t image_phash(IplImage *img) {
+uint64_t image_fingerprint(IplImage *img) {
     int x, y;
     double avg = 0;
     uint64_t phash = 0, phash_mask = 1;
 
-    IplImage *small = cvCreateImage(cvSize(DCT_SIZE, DCT_SIZE), img->depth, img->nChannels);
-    IplImage *mono  = cvCreateImage(cvSize(DCT_SIZE, DCT_SIZE), img->depth, 1);
+    IplImage *mono  = cvCreateImage(cvSize(img->width, img->height), img->depth, 1);
+    IplImage *small = cvCreateImage(cvSize(DCT_SIZE, DCT_SIZE),      img->depth, 1);
 
-    cvResize(img, small, CV_INTER_CUBIC);
-    img->nChannels == 1 ? cvCopy(small, mono, 0) : cvCvtColor(small, mono, CV_RGB2GRAY);
+    img->nChannels == 1 ? cvCopy(img, mono, 0) : cvCvtColor(img, mono, CV_RGB2GRAY);
+    cvResize(mono, small, CV_INTER_CUBIC);
 
-    CvMat *dct = cvCreateMat(DCT_SIZE, DCT_SIZE, CV_32FC1);
-    for (x = 0; x < DCT_SIZE; x++) {
-        for (y = 0; y < DCT_SIZE; y++) {
-            cvSet2D(dct, x, y, cvScalarAll(cvGetMonoPixel(mono, x, y)));
-        }
-    }
+    CvMat *dct = cvCreateMat(DCT_SIZE, DCT_SIZE, CV_64FC1);
+
+    cvConvertScale(small, dct, 1, 0);
+    cvTranspose(dct, dct);
 
     cvDCT(dct, dct, CV_DXT_ROWS);
     cvSet2D(dct, 0, 0, cvScalarAll(0));
 
-    for (y = 0; y < 8; y++) {
-        for (x = 0; x < 8; x++) {
-            avg += cvGet2D(dct, x, y).val[0];
-        }
-    }
-
-    avg /= 63.0;
+    CvMat roi;
+    cvGetSubRect(dct, &roi, cvRect(0, 0, 8, 8));
+    avg = cvAvg(&roi, 0).val[0] * 64.0 / 63.0;
 
     for (x = 7; x >= 0; x--) {
         for (y = 7; y >= 0; y--) {
-            if (cvGet2D(dct, x, y).val[0] > avg) phash |= phash_mask;
+            if (cvGet2D(dct, x, y).val[0] > avg)
+                phash |= phash_mask;
             phash_mask = phash_mask << 1;
         }
     }
 
-    phash = phash & 0x7FFFFFFFFFFFFFFF;
-
     cvReleaseMat(&dct);
     cvReleaseImage(&mono);
     cvReleaseImage(&small);
@@ -117,11 +111,11 @@ IplImage* rb_image_handle(VALUE self) {
   return handle;
 }
 
-VALUE rb_image_hash(VALUE self) {
-    VALUE hash = rb_iv_get(self, "@hash");
+VALUE rb_image_fingerprint(VALUE self) {
+    VALUE hash = rb_iv_get(self, "@fingerprint");
     if (NIL_P(hash)) {
-        hash = SIZET2NUM(image_phash(rb_image_handle(self)));
-        rb_iv_set(self, "@hash", hash);
+        hash = SIZET2NUM(image_fingerprint(rb_image_handle(self)));
+        rb_iv_set(self, "@fingerprint", hash);
     }
     return hash;
 }
@@ -135,10 +129,9 @@ VALUE rb_image_initialize(VALUE self, VALUE file) {
     return self;
 }
 
-
 VALUE rb_image_distance(VALUE self, VALUE other) {
-    VALUE hash1 = rb_image_hash(self);
-    VALUE hash2 = rb_image_hash(other);
+    VALUE hash1 = rb_image_fingerprint(self);
+    VALUE hash2 = rb_image_fingerprint(other);
     int dist    = popcount(NUM2ULONG(hash1) ^ NUM2ULONG(hash2));
     return INT2NUM(dist);
 }
@@ -155,7 +148,7 @@ VALUE rb_image_distance_func(VALUE self, VALUE file1, VALUE file2) {
         rb_raise(rb_eArgError, "Invalid image or unsupported format: %s", CSTRING(file2));
     }
 
-    int dist = popcount(image_phash(img1) ^ image_phash(img2));
+    int dist = popcount(image_fingerprint(img1) ^ image_fingerprint(img2));
 
     cvReleaseImage(&img1);
     cvReleaseImage(&img2);
@@ -163,13 +156,20 @@ VALUE rb_image_distance_func(VALUE self, VALUE file1, VALUE file2) {
     return INT2NUM(dist);
 }
 
-VALUE rb_image_phash_func(VALUE self, VALUE file) {
+VALUE rb_popcount_func(VALUE self, VALUE value) {
+    if (TYPE(value) != T_BIGNUM && TYPE(value) != T_FIXNUM)
+        rb_raise(rb_eArgError, "value needs to be a number");
+    return INT2NUM(popcount(NUM2INT(value)));
+}
+
+
+VALUE rb_image_fingerprint_func(VALUE self, VALUE file) {
     IplImage *img;
     img = cvLoadImage(CSTRING(file), -1);
     if (!img)
         rb_raise(rb_eArgError, "Invalid image or unsupported format: %s", CSTRING(file));
 
-    uint64_t phash = image_phash(img);
+    uint64_t phash = image_fingerprint(img);
     cvReleaseImage(&img);
 
     return SIZET2NUM(phash);
@@ -178,10 +178,11 @@ VALUE rb_image_phash_func(VALUE self, VALUE file) {
 void Init_similie() {
     VALUE cSimilie = rb_define_class("Similie", rb_cObject);
     rb_define_alloc_func(cSimilie, rb_image_alloc);
-    rb_define_method(cSimilie, "initialize", RUBY_METHOD_FUNC(rb_image_initialize), 1);
-    rb_define_method(cSimilie, "hash",       RUBY_METHOD_FUNC(rb_image_hash),       0);
-    rb_define_method(cSimilie, "distance",   RUBY_METHOD_FUNC(rb_image_distance),   1);
+    rb_define_method(cSimilie, "initialize",  RUBY_METHOD_FUNC(rb_image_initialize),  1);
+    rb_define_method(cSimilie, "fingerprint", RUBY_METHOD_FUNC(rb_image_fingerprint), 0);
+    rb_define_method(cSimilie, "distance",    RUBY_METHOD_FUNC(rb_image_distance),    1);
 
-    rb_define_singleton_method(cSimilie, "distance", RUBY_METHOD_FUNC(rb_image_distance_func), 2);
-    rb_define_singleton_method(cSimilie, "phash",    RUBY_METHOD_FUNC(rb_image_phash_func),    1);
+    rb_define_singleton_method(cSimilie, "distance",    RUBY_METHOD_FUNC(rb_image_distance_func),    2);
+    rb_define_singleton_method(cSimilie, "fingerprint", RUBY_METHOD_FUNC(rb_image_fingerprint_func), 1);
+    rb_define_singleton_method(cSimilie, "popcount",    RUBY_METHOD_FUNC(rb_popcount_func),          1);
 }

data/test/helper.rb

@@ -0,0 +1,8 @@
+require_relative '../ext/similie'
+
+require 'minitest/unit'
+require 'minitest/spec'
+
+$testdir = File.absolute_path(File.dirname(__FILE__))
+
+MiniTest::Unit.autorun

data/test/test_basic.rb

@@ -12,12 +12,17 @@ describe 'Similie image load' do
   it 'should hash image' do
     img = Similie.new(File.join($testdir, 'lena1.png'))
     assert img
-    assert_equal 36170087496991428, img.hash
+    assert_equal 36170087496991428, img.fingerprint
   end
 
   it 'should hash image using class method' do
-    hash = Similie.phash(File.join($testdir, 'lena1.png'))
+    hash = Similie.fingerprint(File.join($testdir, 'lena1.png'))
     assert hash
     assert_equal 36170087496991428, hash
   end
+
+  it 'should expose popcount' do
+    assert_equal 3, Similie.popcount(0x03 ^ 0x08)
+    assert_equal 4, Similie.popcount(0x07 ^ 0x08)
+  end
 end

metadata

@@ -4,9 +4,9 @@ version: !ruby/object:Gem::Version
   prerelease: false
   segments: 
   - 0
-  - 2
-  - 6
-  version: 0.2.6
+  - 3
+  - 1
+  version: 0.3.1
 platform: ruby
 authors: 
 - Bharanee Rathna
@@ -14,11 +14,11 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2011-06-08 00:00:00 +10:00
+date: 2011-11-20 00:00:00 +11:00
 default_executable: 
 dependencies: []
 
-description: "\n    Similie does image fingerprinting using discrete cosine transform\n    and similarity comparison using Hamming distance on fingerprints.\n  "
+description: similie is an image fingerprinting & comparison utility
 email: deepfryed@gmail.com
 executables: []
 
@@ -27,12 +27,12 @@ extensions:
 extra_rdoc_files: 
 - README.rdoc
 files: 
-- VERSION
 - ext/similie.c
 - README.rdoc
+- ext/extconf.rb
+- test/helper.rb
 - test/test_basic.rb
 - test/test_distance.rb
-- ext/extconf.rb
 has_rdoc: true
 homepage: http://github.com/deepfryed/similie
 licenses: []
@@ -64,7 +64,6 @@ rubyforge_project:
 rubygems_version: 1.3.7
 signing_key: 
 specification_version: 3
-summary: Compute image fingerprints and similarity
-test_files: 
-- test/test_basic.rb
-- test/test_distance.rb
+summary: compute image fingerprints and similarity
+test_files: []
+

data/VERSION

@@ -1 +0,0 @@
-0.2.6