Research on the Improvement of EPTA Parallel Thinning Algorithm

Abstract

Aiming at the problem of non-smooth contours (contour noises, scratches and jitters), it is easy to generate redundant branches. Based on EPTA algorithm, an improved parallel thinning algorithm is proposed. Through an iteration that eliminates the restrictions and proportional iterations, the pixels in the no-branching phenomenon are first regularly refined; then, the global smoothing refinement is performed to solve the branching problem caused by partial pixels. Experimental results show that the improved algorithm can solve the problem of the existing thinning algorithm, such as poor noise immunity, redundant branches and thin lines, and guarantee the robustness of the thinning results. At the same time, the erosion problem of the original algorithm for 4x4 square is also solved.

Skeleton extraction and pose estimation of piglets using ZS-DLC-PAF

“…However, the parity of the pixel coordinate index values will directly affect the results. To increase speed, a general parallel computing platform and programming model were introduced to accelerate the algorithm [13] . Li et al [14] pointed out that an uneven surface contour will lead to a problem in the thinning results, i.e., the bifurcation outline of the image will be affected by noise.…”