Study On Some Key Technologies In Image Signal Processor

Image Signal Processor (ISP), as a core part of digital camera system, is responsible for a variety of image processing functions during the process of shooting. ISP system is currently facing two major challenges: 1. As sensor size increases, LineBuffer size to become a major bottleneck in system integration. How to reduce LineBuffer related resource and reduce ISP costs becomes an important issue to improve the performance ISP. 2. How to improve the color filter interpolation (CFA) algorithm to meet ever-increasing visual requirements.In this paper, an improved LineBuffer model is firstly proposed. Using this model, LineBuffer can be completed in to single-port ram rather than dual-port ram and complexity of control logic of ram is thus reduced. Then, section based data path and block based data path are proposed. Data paths are compared in LineBuffer consumed, SDRAM read and write operation times, the proportion of redundant data and听the controlling complexity. Block-based data path achieved a good balance in SDRAM and LineBuffer, which is a very good data path solution. By adopting improved LineBuffer model and block-based data path, LineBuffer related resource can be greatly decreased.Finally, an improved edge adaptive color demosaicking method is presented. Edges are detected by an improved edge indicator, and missing color values are estimated in chosen direction. In estimation stage, weighted average of Taylor series is adopted. Proposed CFA algorithm not only outperforms other existed methods in subjective visual quality but also in terms of peak signal-to-noise ratio (PSNR).