Fig. 3From: A learned sparseness and IGMRF-based regularization framework for dense disparity estimation using unsupervised feature learningFilters learned at first and second layers of deep deconvolutional network. a Number of filters learned at first layer are 9. b Number of filters learned at second layer are 81 where 36 filters in pair are shown in color and remaining 9 filters are shown as gray scaleBack to article page