28.3: Dense Disparity Map Calculation from Color Stereo Images using Edge Information
Ja Seung Ku, Hui Nam, Chan Young Park, Beom Shik Kim, Yeon‐Gon Mo, Hyoung Wook Jang, Hye‐Dong Kim, Ho Kyoon Chung · SID Symposium Digest of Technical Papers · 2007
Abstract We have developed the stereo‐corresponding algorithm using edge information. The conventional stereo‐corresponding algorithm, SAD (Sum of Absolute Difference), has a good quality in case of texture region, but it has a false matching in the non‐texture region. In order to reduce the false matching in the non‐texture region, the new cost function, which is defined as SED (Sum of Edge Difference) based on the global optimization, is added to the cost function of SAD. We evaluate our algorithm and benchmark Middlebury database. The experimental results show that our proposed algorithm successfully produces piecewise smooth disparity maps while reducing the false matching in the non‐texture region. Moreover, our algorithm is faster in getting the best quality than SAD.