Modification of Pixel-swapping Algorithm with Initialization fr om a Sub-pixel/pixel Spatial Attraction Model

Pixel-swapping algorithm is a simple and efficient technique for sub-pixel mapping (Atkinson, 2001 and 2005). It was initially applied in shoreline and rural land-cover mapping but has been expanded to other land-cover mapping. However, due to its random initializing process, this algorithm must swap a large number of sub-pixels, and therefore it is computation intensive. This computing power consumption intensifies when the scale factor is large. A new, modified pixel-swapping algorithm (MPS) is presented in this paper to reduce the computation time, as well as to improve sub-pixel mapping accuracy. The MPS algorithm replaces the original random initializing process with a process based on a sub-pixel/pixel spatial attraction model. The new algorithm was used to allocate multiple land-covers at the subpixel level. The results showed that the MPS algorithm outperformed the original algorithm both in sub-pixel mapping accuracy and computational time. The improvement is especially significant in the case of large scale factors. Furthermore, the MPS is less sensitive to the size of neighboring sub-pixels and can still result in increased accuracy even if the size of neighbors is small. The MPS was also much less time consuming, as it reduced both the iterations and total amount of swapping needed. Introduction Since the launch of the first Earth observation satellite, remote sensing imagery has been utilized increasingly in many applications including land-cover analysis, environmental monitoring, mineral exploration, military surveillance, etc. A common problem associated with the application of satellite images, however, is the frequent occurrence of mixed pixels (e.g., Foody, 2004). Mixed pixels in traditional land-use Modification of Pixel-swapping Algorithm with Initialization fr om a Sub-pixel/pixel Spatial Attraction Model Zhangquan Shen, Jiaguo Qi, and Ke Wang and land-cover classification processes are often classified into a single land-use/land-cover type, without preserving the information that the pixel contains a mixture of multiple land-uses and covers. Soft classification techniques were introduced to avoid the loss of information by assigning a pixel to multiple land-use/land-cover classes according to the area each use/cover represents within the pixel. This soft classification technique generates a number of fractional images equal to the number of classes (Mertens et al., 2003). Unfortunately, the results from soft classification do not specify the location of each class within that particular pixel. In many practical applications, it is often desirable to know where each class is located within the pixel, in order to obtain detailed spatial patterns of land-use and land-cover. Sub-pixel mapping (or super-resolution mapping) was then introduced (Atkinson et al., 1997; Atkinson, 1997) to achieve this desirable goal, using the information obtained from both soft and hard classification techniques. The aim of sub-pixel mapping is to determine the most suitable locations for different classes produced from a soft classification. It attempts to allocate each thematic mapping fraction to an appropriate subpixel location using soft classification results. Hence, sub-pixel mapping is a spatial allocation technique that transforms a soft classification into a finer scale hard classification. Several approaches have been proposed to tackle the sub-pixel mapping issue: pixel swapping (Atkinson, 2001 and 2005), image sharpening (Foody, 1998; Gross and Schott, 1998), knowledge-based analysis (Schneider, 1993), Hopfield neural networks (Tatem et al., 2001 and 2002), de-convolution filters (Pinilla and Ariza, 2002), linear optimization (Verhoeye et al., 2002), genetic algorithms (Mertens et al., 2003), feedforward neural networks (Mertens et al., 2004), Markov random field-based approach (Kasetkasem et al., 2005), algorithm based on sub-pixel/pixel spatial attraction models (Mertens et al., 2006), and integration of information from indicator co-kriging or indicator kriging (Boucher and Kyriakidis, 2006 and 2007). One of the sub-pixel mapping algorithms, named pixelswapping (PS), was first proposed by Atkinson (2001 and 2005) and tested with synthesized images. Due to its simplicity and efficiency, the PS algorithm has been used successfully for mapping shorelines in Malaysia (Muslim et al., 2006) and rural land-cover features in the Christchurch area of PHOTOGRAMMETRIC ENGINEER ING & REMOTE SENS ING May 2009 557 Zhangquan Shen is with the College of Environmental and Resource Sciences, Zhejiang University, Hangzhou, 310029, P. R. China, and previously with the Center for Global Change and Earth Observations, Michigan State University, East Lansing, MI 48823 ([email protected]). Jiaguo Qi is with the Center for Global Change and Earth Observations and Department of Geography, Michigan State University, East Lansing, MI 48823, and the Institute of Geographic Sciences and Natural Resources, Chinese Academy of Science, Beijing, China. Ke Wang is with the College of Environmental and Resource Sciences, Zhejiang University, Hangzhou 310029, P. R. China. Photogrammetric Engineering & Remote Sensing Vol. 75, No. 5, May 2009, pp. 557–567. 0099-1112/09/7505–0557/$3.00/0 © 2009 American Society for Photogrammetry and Remote Sensing 557-567_07-105.qxd 4/16/09 10:44 AM Page 557