Enhancing the Locational Perception of Soft Classified Satellite Imagery Through Evaluation and Development of the Pixel Swapping Technique

Spatial component is the key and most likely the first element of map making so that accurate spatial information improves the locational perception of map users. In this regard, soft classified satellite imagery conveys class proportions within pixels; however spatial distribution of the sub-pixels remains unknown. So, different visualization techniques (e.g. pie-chart representation of the proportions) are suggested to communicate the detailed land cover information. However, in each of which, the perception of actual spatial location of sub-pixels is definitely difficult for map users. Recently, the Super Resolution Mapping (SRM) techniques have been developed for optimization of the sub-pixels spatial arrangement based on the concepts of spatial dependency. These are relatively new methods which a comprehensive study on their performance and also their decisive parameters is a central issue for sub-pixel land cover mapping. In this research, the binary Pixel Swapping (PS) algorithm, as a prominent SRM algorithm, is developed for multivariate land cover mapping and the accuracy of the proposed method is evaluated in two procedures of independent and dependent of the soft classification error. Likewise, the impact of some parameters (e.g. zoom factor, neighborhood level and weighting function) is investigated on the efficiency of the algorithm. According to the results, the overall accuracy of the PS technique is extremely dependent on the accuracy of its input data (outputs of the soft classification). Furthermore, as a key result of this chapter, it is indicated that by increasing the zoom factor, the overall accuracy of the algorithm decreases. Also, the second level of neighborhood and inverse/square inverse distance functions has demonstrated the highest accuracies. Considering lower values than 5 for zoom factor, overall accuracy of the algorithm is determined higher than 90 % in procedure of optimizing the sub-pixels spatial arrangement.