SPOT 5 Supermode

Supermode is a unique image sampling process that offers significantly improved resolution in the panchromatic band without major modifications to the satellite, thanks to sophisticated processing operations on the ground.

The Supermode process generates a single image at a resolution of 2.5 metres from two panchromatic images acquired simultaneously at a resolution of 5 metres and offset vertically and horizontally by 2.5 metres.

Onboard processing
Imagery is acquired by two dedicated arrays of CCD detectors offset in the focal plane. The instrument thus generates two 5-metre images that are processed independently by the payload.

Ground processing
Final Supermode products are generated in three steps: interpolation, deconvolution and noise removal.

Interpolation
The first step consists in interlacing the two images acquired by the offset arrays and interpolating "missing" pixels to obtain an image that is twice as sharp:

 

Deconvolution
The second step compensates for blurring introduced by the instrument, which is strong at high frequencies since Shannon's condition is almost satisfied. This operation, called deconvolution, applies a filter representing the instrument's inverse transfer function.

Deconvolution amplifies noise at high frequencies in the image, so a third processing step is required to remove noise.

Noise removal
This last step reduces the noise in the image to an acceptable preset level. It is also the most complex, since it involves about 1,500 operations (compared to 400 for the other steps). The algorithm is based on a method called fixed chosen noise restoration (FCNR), which aims to control noise removal in the most sensitive parts of the image-those that are virtually uniform. It achieves this by shrinking wavelet packet coefficients in non-linear fashion, employing joint-adaptive space and frequency wavelet packet decomposition. It is thus able to distinguish uniform areas in the image while factoring in noise variations due to the signal, and noise amplification, which is frequency-dependent via the deconvolution function. Noise removal therefore consists in thresholding noisy wavelet coefficients in the image at different image restoration levels.

This three-step process produces a single 2.5-metre black-and-white image of 24,000 x 24,000 pixels from two 5-metre images of 12,000 x 12,000 pixels.

A learning process
Noise is removed by learning from a noisy image (in fact, a simulated uniform landscape viewed by the instrument). The process consists in obtaining a target output noise level equivalent to the noise in the image before the deconvolution step. This is achieved by calculating the attenuation coefficients to be applied to the wavelet coefficients. The attenuation coefficients applied when removing noise from a real image are proportional to those calculated during the learning phase.

 

Marseille harbour and sea wall(SPOT 5 simulated Supermode image at 2.5 m)

 

Other simulated SPOT 5 Supermode 2.5-metre images:

Marignane
Bouc harbour
Lacaune forest
Highway interchange

Supermode applications
Supermode will serve a broad spectrum of SPOT data applications, particularly forest inventorying, agriculture, surveillance and natural hazard mitigation.

Forest inventorying
Forest mapping is designed to inventory plant species and evaluate timber resources at national scales. Today, SPOT data are used increasingly to update maps. SPOT 5 Supermode will provide new information on land cover and canopy closure, and detect crown thinning, clear-cuts, strips and vertical stratification. Supermode's enhanced resolution could be used to add more detail to forest maps.

SPOT 5 simulated image - forests in the Hérault region of France

 

Agriculture
SPOT panchromatic data are used at the start of agricultural survey campaigns to identify land cover and measure field acreages. Crops are identified by analysing multispectral data acquired at different dates. Supermode data will make it possible to discriminate closely aligned crops such as orchards and vineyards, which are difficult to identify in currently available satellite imagery. They will also yield highly detailed information about crop health and status within individual parcels.

SPOT 5 simulated image - farmland (near Lavaur, south-west France)

 

Surveillance and hazard mitigation
SPOT 5 Supermode 2D or stereoscopic data are ideally suited to reconnaissance and interpretation for mission planning, strategic intelligence, crisis monitoring, and monitoring of disarmament accords and industrial and urban sites.

Perspective view of Fos-sur-Mer, France (2.5-metre Supermode image)