The digital image filtering method for noise reduction and edge preservation, along with a specialized processing circuit for pipeline processing, is used to implement this method.

Abstract

The invention relates to a method for edge-preserving noise reduction in image filtering and a corresponding dedicated processing circuit to implement this method. In the method described in the invention, for each central pixel, a neighborhood window is established on the input image and the navigation image. The edge costs between adjacent pixels in the navigation image are determined and used to calculate the local geodetic cost from the central pixel to neighboring pixels using a two-pass wave propagation procedure with a defined update schedule. The geodetic cost is used to generate geodetic weights, which are then combined with spatial weights and photometric weights to create the final filter weights. The corresponding circuit consists of a window generator, an edge cost calculation block, a geodetic propagation block, a weight generator block, a weighted synthesis block, and an output normalization block organized in a pipeline. The invention allows for the exploitation of interconnected geodetic-type information while maintaining a steady processing flow, defined latency, and high throughput for real-time image processing.

Images