Add the two missing deconvolution tests.

diff --git a/deconvolution_sharpen_effect_test.cpp b/deconvolution_sharpen_effect_test.cpp
index bf6dfb81b8f105dbbf7d5d0a96f9d8b5df5f9c86..ec78968dad3cb25798f93bf99da506be7680ef80 100644 (file)
--- a/deconvolution_sharpen_effect_test.cpp
+++ b/deconvolution_sharpen_effect_test.cpp
@@ -4,6 +4,29 @@
 #include "gtest/gtest.h"
 #include "deconvolution_sharpen_effect.h"
 
+TEST(DeconvolutionSharpenEffect, IdentityTransformDoesNothing) {
+       const int size = 4;
+
+       float data[size * size] = {
+               0.0, 1.0, 0.0, 1.0,
+               0.0, 1.0, 1.0, 0.0,
+               0.0, 0.5, 1.0, 0.5,
+               0.0, 0.0, 0.0, 0.0,
+       };
+       float out_data[size * size];
+
+       EffectChainTester tester(data, size, size, FORMAT_GRAYSCALE, COLORSPACE_sRGB, GAMMA_LINEAR);
+       Effect *deconvolution_effect = tester.get_chain()->add_effect(new DeconvolutionSharpenEffect());
+       ASSERT_TRUE(deconvolution_effect->set_int("matrix_size", 5));
+       ASSERT_TRUE(deconvolution_effect->set_float("circle_radius", 0.0f));
+       ASSERT_TRUE(deconvolution_effect->set_float("gaussian_radius", 0.0f));
+       ASSERT_TRUE(deconvolution_effect->set_float("correlation", 0.0001f));
+       ASSERT_TRUE(deconvolution_effect->set_float("noise", 0.0f));
+       tester.run(out_data, GL_RED, COLORSPACE_sRGB, GAMMA_LINEAR);
+
+       expect_equal(data, out_data, size, size);
+}
+
 TEST(DeconvolutionSharpenEffect, DeconvolvesCircularBlur) {
        const int size = 13;
 
@@ -102,5 +125,72 @@ TEST(DeconvolutionSharpenEffect, DeconvolvesGaussianBlur) {
        expect_equal(expected_data, out_data, size, size);
 }
 
-// TODO: Test no-op (both radii equal to zero).
-// TODO: Test correlation and noise parameters.
+TEST(DeconvolutionSharpenEffect, NoiseAndCorrelationControlsReduceNoiseBoosting) {
+       const int size = 13;
+       const float sigma = 0.5f;
+
+       float data[size * size], out_data[size * size];
+       float expected_data[size * size] = {
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 1.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+               0.0, 0.0, 0.0, 0.0, 0.000000, 0.000000, 0.000000, 0.000000, 0.000000, 0.0, 0.0, 0.0, 0.0, 
+       };
+
+       // Gaussian kernel.
+       float sum = 0.0f;       
+       for (int y = 0; y < size; ++y) {
+               for (int x = 0; x < size; ++x) {
+                       float z = hypot(x - 6, y - 6);
+                       data[y * size + x] = exp(-z*z / (2.0 * sigma * sigma)) / (2.0 * M_PI * sigma * sigma);
+                       sum += data[y * size + x];
+               }
+       }
+       for (int y = 0; y < size; ++y) {
+               for (int x = 0; x < size; ++x) {
+                       data[y * size + x] /= sum;
+               }
+       }
+
+       // Corrupt with some uniform noise.
+       srand(1234);
+       for (int i = 0; i < size * size; ++i) {
+               data[i] += 0.1 * ((float)rand() / RAND_MAX - 0.5);
+       }
+
+       EffectChainTester tester(data, size, size, FORMAT_GRAYSCALE, COLORSPACE_sRGB, GAMMA_LINEAR);
+       Effect *deconvolution_effect = tester.get_chain()->add_effect(new DeconvolutionSharpenEffect());
+       ASSERT_TRUE(deconvolution_effect->set_int("matrix_size", 5));
+       ASSERT_TRUE(deconvolution_effect->set_float("circle_radius", 0.0f));
+       ASSERT_TRUE(deconvolution_effect->set_float("gaussian_radius", 0.5f));
+       ASSERT_TRUE(deconvolution_effect->set_float("correlation", 0.5f));
+       ASSERT_TRUE(deconvolution_effect->set_float("noise", 0.1f));
+       tester.run(out_data, GL_RED, COLORSPACE_sRGB, GAMMA_LINEAR);
+
+       float sumsq_in = 0.0f, sumsq_out = 0.0f;
+       for (int i = 0; i < size * size; ++i) {
+               sumsq_in += data[i] * data[i];
+               sumsq_out += out_data[i] * out_data[i];
+       }
+
+       // The limits have to be quite lax; deconvolution is not an exact operation.
+       // We special-case the center sample since it's the one with the largest error
+       // almost no matter what we do, so we don't want that to be the dominating
+       // factor in the outlier tests.
+       int center = size / 2;
+       EXPECT_GT(out_data[center * size + center], 0.5f);
+       out_data[center * size + center] = 1.0f;
+       expect_equal(expected_data, out_data, size, size, 0.20f, 0.005f);
+
+       // Check that we didn't boost total energy (which in this case means the noise) more than 10%.
+       EXPECT_LT(sumsq_out, sumsq_in * 1.1f);
+}