Research On Lapped Transform And Vector Quantization Based Image Coding Algorithm And Applications

As a carrier of information, image has increasing applications in daily life as well as in industry. But it is difficult to transmit and process digital images because of the huge amount of data. As a result, various kinds of image compression methods have been proposed to remove the redundancy in image. However, most of them could not give good performance in a high compression ratio situation, and always have a high computational complex, which make it hard to implement in resource limited and real-time system. To solve this problem, this paper addressed the algorithms on lapped transforms and vector quantization, which have good performance under high compression ratio. After that, we implemented the algorithm on a PDA embedded system.Lapped transforms is a kind of transform which is in competition with DCT and wavelet. In this paper, we studied the theory, background, and definition of the lapped transform. And then, some typical state-of-the-art lapped transforms were simulated. In order to compare with lapped transform, we also simulated the DCT and wavelet transforms. Consider that SPIHT algorithm can present a good performance in the frequency domain coding, we implemented the SPIHT algorithm and combined it with the all kinds of transforms mentioned above, which can evaluate the capability of coding of the various transforms. Through the comparisons in simulations, this paper illustrated the efficiencies and characters of all kinds of transforms coding.In order to make compression ratio higher, in this paper, we also introduced vector quantization, and investigated an improved multistage vector quantization, which employ a classifier in codebook design. The traditional full search vector quantization (FSVQ) can give a high quality in image coding, but it is time and memory space consuming. Hence, the Multistage Vector Quantization (MSVQ) that has much lower computation and memory requirements is proposed. Nevertheless, MSVQ can not give the same performance in the reconstruct quality of image coding. Hence, we proposed a classified multistage vector quantization (CMSVQ) to keep low computational complex and enhance the quality of image coding. Simulations show that the CMSVQ scheme can achieve a better trade-off between image quality and computational complex.At last of our paper, we also present a combination scheme between lapped transform and classified multistage vector quantization. The results show that the proposed scheme can increase compression ratio and decease computational and space complexities effectively. After that, we make the algorithm implement on a PDA embedded systems. Simulation result shows that the scheme mentioned above can give a high performance in a resources limited hardware platform.