ICALAB for Image Processing - benchmarks |
The directory BENCHMARKS consists of a collection of color and
grayscale images with various resolutions.
The images are saved individually and all collected in one MATLAB file
NAME.mat in order to save time during loading.
You can download all benchmarks in matlab format (*mat), files tared and gzipped to one big file: benchmarks_2d.tgz (4.42MB).
The most interesting benchmarks are briefly described below.
This color and grayscale benchmark contains 12 two-dimensional sine wave sources as illustrated in Fig. 1B.
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| color images | sin2d_128x128_color.mat | sin2d_128x128_color.mat.gz | sin2d_128x128_color.tgz |
| grayscale images | sin2d_128x128_gray.mat | sin2d_128x128_gray.mat.gz | sin2d_gray_128x128.tgz |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
These color and grayscale benchmarks are similar to the sin2d benchmarks, but the 2-D sinewaves are full wave rectified (see Fig. 4B).
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| color images | FWRsin2d_128x128_color.mat | FWRsin2d_128x128_color.mat.gz | FWRsin2d_128x128_color.tgz |
| grayscale images | FWRsin2d_128x128_gray.mat | FWRsin2d_128x128_gray.mat.gz | FWRsin2d_gray_128x128.tgz |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
These color and grayscale benchmarks are similar to the benchmarks sin2d, but the two-dimensional sine waves are half-wave rectified so the images become more sparsely distributed (see Fig. 5B).
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| color images | HWRsin2d_128x128_color.mat | HWRsin2d_128x128_color.mat.gz | HWRsin2d_128x128_color.tgz |
| grayscale images | HWRsin2d_128x128_gray.mat | HWRsin2d_128x128_gray.mat.gz | HWRsin2d_gray_128x128.tgz |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
These benchmarks are similar to the above benchmarks, but sine waves are one-dimensional (see Figs 6B, 7B, 8B).
| type | mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| sin1d | color images | sin1d_128x128_color.mat | sin1d_128x128_color.mat.gz | sin1d_128x128_color.tgz |
| grayscale images | sin1d_128x128_gray.mat | sin1d_128x128_gray.mat.gz | sin1d_gray_128x128.tgz | |
| full-rectified sin1d | color images | FWRsin1d_128x128_color.mat | FWRsin1d_128x128_color.mat.gz | FWRsin1d_128x128_color.tgz |
| grayscale images | FWRsin1d_128x128_gray.mat | FWRsin1d_128x128_gray.mat.gz | FWRsin1d_gray_128x128.tgz | |
| half-rectified sin1d | color images | HWRsin1d_128x128_color.mat | HWRsin1d_128x128_color.mat.gz | HWRsin1d_128x128_color.tgz |
| grayscale images | HWRsin1d_128x128_gray.mat | HWRsin1d_128x128_gray.mat.gz | HWRsin1d_gray_128x128.tgz |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
| 3D view | color image | grayscale image |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
The benchmarks edges1d, edges1ds and edges2d
consist of grayscale level "sparse" images
representing regular edges. Please try any HOS ICA algorithm like SANG,
JADE or FICA to extract such images from their mixtures.
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| grayscale images | edges1d.mat | edges1d.mat.gz | edges1d.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| grayscale images | edges1ds.mat | edges1ds.mat.gz | edges1ds.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| grayscale images | edges2d.mat | edges2d.mat.gz | edges2d.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
This benchmark consists of 10 natural images (e.g. faces, animals, objects) and two synthetically created noise images.
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| color images | natural_images_128x128_color.mat | natural_images_128x128_color.mat.gz | natural_images_128x128_color.tgz |
| grayscale images | natural_images_128x128_gray.mat | natural_images_128x128_gray.mat.gz | natural_images_gray_128x128.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
This benchmark consists of 16 typical textures.
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| color images | textures_128x128_color.mat | textures_128x128_color.mat.gz | textures_128x128_color.tgz |
| grayscale images | textures_128x128_gray.mat | textures_128x128_gray.mat.gz | textures_gray_128x128.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
This benchmark consists of 9 human faces selected from face databases of Psychology Department of University of Stirling
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| grayscale images | faces_color.mat | faces_color.mat.gz | faces_color.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
This benchmark consists of 6 human faces selected from face databases of Psychology Department of University of Stirling
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| grayscale images | faces.mat | faces.mat.gz | faces.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
These benchmarks consist each of 10 typical MRI images selected from databases of The MRI Tutor Web Site and The Whole Brain Atlas
| mode | matlab files (*.mat) | gzipped matlab files (*.mat.gz) | tared and gzipped tiff files (*.tgz) |
| color images | MRI_color.mat | MRI_color.mat.gz | MRI_color.tgz |
| grayscale images | MRI_gray.mat | MRI_gray.mat.gz | MRI_gray.tgz |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |