Dense matching for WorldView-3 multispectral stereo images

I. Ling Kuo, Tee-Ann Teo

Research output: Contribution to conferencePaperpeer-review


Due to the development of spatial resolution, spectral resolution and multi-image tasking of high-resolution satellite images, the generation of digital surface model (DSM) from high-resolution satellite images becomes an important and practical application. Nowadays, different image matching approaches such as global matching and semi-global matching has been developed to obtain dense point clouds from multi-view images. The image-derived and lidar-derived dense point clouds are the two competitive products. Although LiDAR may provide reliable points, the multi-view image matching is more cost effective than lidar. The conventional semi-global matching (SGM) is mainly applied for center perspective frame images. SGM is rarely applied for linear pushbroom satellite stereo images. Besides, the spectral image is usually converted to intensity blind image in image matching. The spectral information might be a useful information in image matching for complex scene. This study develops a SGM for WorldView-3 high resolution satellite images. The research compares two different high density matching strategies: (1) RFM-based SGM for panchromatic stereo image, and (2) RFM-based SGM for multi-spectral stereo images. The RFM-based SGM utilizes quasi-epipolar images from rational function in matching cost computation. The RFM-based SGM compares the results using different spectral bands. The experiment images contain panchromatic and 8 band multi-spectral stereo images from WorldView-3 satellite. The experimental results indicated that the NIR band generated more reliable points than panchromatic in vegetated area. The selection of multispectral information for dense matching may improve the point density for different land covers.

Original languageEnglish
StatePublished - 1 Jan 2015
Event36th Asian Conference on Remote Sensing: Fostering Resilient Growth in Asia, ACRS 2015 - Quezon City, Metro Manila, Philippines
Duration: 24 Oct 201528 Oct 2015


Conference36th Asian Conference on Remote Sensing: Fostering Resilient Growth in Asia, ACRS 2015
CityQuezon City, Metro Manila


  • Dense matching
  • Multispectral image
  • Stereo
  • Worldview-3

Fingerprint Dive into the research topics of 'Dense matching for WorldView-3 multispectral stereo images'. Together they form a unique fingerprint.

Cite this