Simulation of 1KM time series NDVI from 8km time series NDVI image

1. wavelet transform

Today, the use of wavelet has become pervasive in mathematics, physics, digital signal/image processing, and geophysics. The wavelet transform leads to the concept of multi-scale analysis, where images are decomposed into structures and then analyzed at successive scales (or spatial resolutions). Just like the two dimensional discrete Fourier transform decomposes an image into a weighted sum of global cosine and sine functions, when the wavelet transform is applied, an image is decomposed into three detailed images and an approximation image. If the transform is continued, the approximation image is in turn decomposed further into three detailed images and another approximation image.

The respective summed weights of the wavelets are called the wavelet coefficients. Wavelet coefficients are a measure of the intensity of the local variations of the image for that individual scale. The value of a coefficient for a particular location at any scale can be understood as a characterization of the image structure at a chosen scale.

decomp1.jpg (14620 bytes)

 

decomp2.jpg (104380 bytes)

 

2. simulation of 1km NDVI from 8km NDVI

When having a series of coarse resolution images and one fine resolution data set, one would be able to simulate fine resolution images for all dates by using wavelet transform.

Currently, 8km NDVI time series from 1982 to 1999 and 1km NDVI for 1985 are available. We used wavelet transform to successfully simulate 1km NDVI time series from 1982 to 1999. These simulated 1km NDVI will be used in the land cover performance project.

decomp3.jpg (29838 bytes)

 

 

1982
F92.bmp (410934 bytes) N82.bmp (410934 bytes)
1988
F88.bmp (410934 bytes) N88.bmp (410934 bytes)
1993
F93.bmp (410934 bytes) N93.bmp (410934 bytes)
1995
F95.bmp (410934 bytes) N95.bmp (410934 bytes)
1999
F99.bmp (410934 bytes) N99.bmp (410934 bytes)