{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

Remote Sensing - a tool for environmental observation

# Transformation functions look like u = f(x,y v =

This preview shows pages 69–73. Sign up to view the full content.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Transformation functions look like: u = f (x,y) v = g (x,y) The (x,y) coordinates describe the position of the GCPs on the map, (u,v) describe the coordinates of the GCPs in the image in terms of rows and columns. f and g are the transform- ation functions. For satellite imagery a first order geometric transformation is best to correct satellite images, airborne images sometimes require higher order transformations. However, higher order transformations require more GCPs. The accuracy of the transformation function is mostly given in RMS-error (root mean square error) and refers to the distance between the input location of a GCP and the retransformed location for the same GCP. Figure 5.7 The six most important flight parameters of an aircraft (Buiten & Clevers 1994). 69 Figure 5.8 Geometric distortions of airborne remote sensing images by aircraft flight movements. Resampling Having determined the geometric transformation function by using GCPs the next step is to compute DN values for the new defined image grid (x,y) based on the DN values in the old grid (rows, columns). The spacing of the grid is chosen according to the pixel size required e.g. from 30 by 30 m to 50 by 50 m. As the old pixels will never fit exactly on the newly defined pixels an interpolation of the new DN values is necessary. This process is called resampling and comprises three techniques (figure 5.9): 1. Nearest neighbour resampling: the new pixel value is the nearest neighbour in the old grid. The advantage of this method is that it is the easiest and fastest method and that the original data values are preserved. 2. Bilinear interpolation: the new pixel value is based upon the distances between the new pixel location and the four closest pixel values in the old grid (inverse distance interpolation). A disadvantage of this method is that the new DN-values are smoothed. 3. Cubic convolution: the new pixel value is computed from the 16 (or even 25) pixel values in the old grid closest to the new location. Lillesand and Kiefer (2000) show the effect of the different resampling methods on a MSS image on page 476. Note the ‘stair stepped’ effect in the nearest neighbour resampling and the smoothing effect in the two other methods. 70 A. B. C. D. Figure 5.9 Resampling techniques: A) Open circles indicate reference grid from input image (and location determined from topographical map or GPS). B) Nearest Neighbourhood resampling: each estimated (new) value receives the value of the nearest point on the reference grid. C) Bilinear interpolation: each estimated value in the output image is formed by calculating the weighted average (inverse distance) of the four neighbours in the input image. D) Cubic convolution: each estimated value in the output matrix is found by computing values within a neighbourhood of 16 pixels in the input image. 71 5.5 Image Enhancement Image enhancement is the modification of a digital image to alter its impact on the viewer....
View Full Document

{[ snackBarMessage ]}

### Page69 / 116

Transformation functions look like u = f(x,y v = g(x,y...

This preview shows document pages 69 - 73. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online