The Research And Implementation On Image Acquisition And Region Of Interest Coding Based On DSP

In recent years, with the extensive application of embedded image acquistion, such as image monitor, medical equipment, industrial automation, intelligent traffic, etc, the requirement of image quality has been improved greatly. The vast amount of image data raises more requirement and challenges to the image acquisition and processing technology. The thesis focuses on the analysis of this issue which is the conflict between image compression efficiency and image quality after reconstruction. To deal with the images in the region of interest and background region with different strategy is to achieve high compression and, meanwhile, to ensure that the image quality of image quality after reconstruction. It's a good solution to the contradiction and a good reference for the related applications and research.Firstly, this thesis introduces the status quo of the subject at china and abroad. Then it analyzes and compares their development and accomplishment. Furthermore, it carries out a detailed analysis of the target system which includes three main characteristics. Then through the comparison of common features of different digital signal processor, it chooses the Blackfin DSP processing system as the core framework. Based on the system requirements, it designs the overall hardware platform architecture and the transplantation ofμCLinux embedded software operating system.Through the analysis of specific characteristics of different types of image sensors, the OV9650 image sensor is selected. Combined the characteristics of BF533 DSP, thesis designs the interface between them, and completes the specific configuration of image sensor. Moreover, in accordance with the unique PPI structure of Blackfin, it achieves the design of rapid image transmission. Finally, all the operations are transplanted to the module driver in embedded operating system platform.Besides, it summarizes and analyses about a variety of edge detection and segmentation algorithms, which mainly include the edge-based operator and the wavelet transform-based operator. Then, depends on the actual demand of the system, it designed the algorithm that utilizes the SUSAN operator for image segmentation and then proceed to the edge tracking and filtering for region of interest extraction. Through the test data and images, its efficiency and result is confirmed.After the completion of image acquisition and image segmentation, it analyses the principles of common image compression coding algorithms and their advantages and disadvantages. Then it selects the wavelet difference reduction algorithm which meets the need of both the efficiency and complexity of image compression coding. In order to achieve region-of-interest based image compression, it also selects the appropriate wavelet basis. Furthermore, by combining the ROI coding priority, it achieves the WDR-ROI coding algorithm. Finally, the test results prove the realization of the proposed algorithm. It accomplishes the balance between the compression ratio and reconstructed image quality, improves the coding efficiency, and saves the data storage space.