DCTlab is a Matlab GUI for performing the steps of DCT based image compression and for assessment of reconstructed image quality. Main program window is divided in parts representing the steps in JPEG image compression (transform, quantisation, coding). Additional menus are available with right-click over displayed image (e.g. for the original image the choices are: Zoom View (shows enlarged picture), Save (saves picture as .bmp file), Assign to workspace (saves image as a variable in Matlab workspace)).
If the selected size of DCT block is 8×8 (as in JPEG), quantisation table and coding can be applied, leading to the compressed bit-stream (no actual bit-stream is generated, just the resulting bit rate is computed). The compression ratio can be adjusted by changing the quantisation matrix (by multiplying it with a selected constant). Zig-zag scanning order, quantisation matrix and Huffman codes are user-definable. The predefined tables, taken from the JPEG standard, come with DCTlab: Qdefault.txt (quantization matrix), Zdeafult.txt (zig-zag pattern), Hdefault (Huffman tables). The file DCTlabDefaults.txt is a 'registry' file pointing to the files containing these tables, so new ones can easily be added in this way. Supported image formats are greyscale .bmp, .png and .mat, and image dimensions are restricted to be 2n×2n.
Required toolbox: Quality Assessment
The main part of the program was developed during my BSc and MSc studies at the Department of Radiocommunications and Microwave Engineering (Video Communications Laboratory - VCL), Faculty of Electrical Engineering and Computing, University of Zagreb.
Functions
This GUI includes the following standalone functions:
- blkdct2cos - 2D DCT on blocks (using cosine function), optimised for square matrices
- blkdct2fft - 2D DCT on blocks (using fft), optimised for square matrices
- blkidct2cos - 2D IDCT on blocks (using cosine function), optimised for square matrices
- blkidct2fft - 2D IDCT on blocks (using fft), optimised for square matrices
- dct2sq - 2D DCT optimised for square matrices
- dctbasis - Computes the elementary basis functions for 1D and 2D DCT
- idct2sq - 2D IDCT optimised for square matrices
Screenshot
