// Unit tests for DeconvolutionSharpenEffect.
-#include "test_util.h"
-#include "gtest/gtest.h"
+#include <math.h>
+#include <stdlib.h>
+
#include "deconvolution_sharpen_effect.h"
+#include "effect_chain.h"
+#include "gtest/gtest.h"
+#include "image_format.h"
+#include "test_util.h"
+
+TEST(DeconvolutionSharpenEffectTest, 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);
-TEST(DeconvolutionSharpenEffect, DeconvolvesCircularBlur) {
+ expect_equal(data, out_data, size, size);
+}
+
+TEST(DeconvolutionSharpenEffectTest, DeconvolvesCircularBlur) {
const int size = 13;
// Matches exactly a circular blur kernel with radius 2.0.
expect_equal(expected_data, out_data, size, size, 0.15f, 0.005f);
}
-TEST(DeconvolutionSharpenEffect, DeconvolvesGaussianBlur) {
+TEST(DeconvolutionSharpenEffectTest, DeconvolvesGaussianBlur) {
const int size = 13;
const float sigma = 0.5f;
expect_equal(expected_data, out_data, size, size);
}
-// TODO: Test no-op (both radii equal to zero).
-// TODO: Test correlation and noise parameters.
+TEST(DeconvolutionSharpenEffectTest, 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);
+}