Image quality

Image properties

Resolution

The image resolution is pending on the camera sensor and its amount of pixels. It is described in horizontal and vertical pixels, e.g., 8820 pix * 5668 pix, which result in 50 Megapixels in Summary.

The physical size of one pixel determines the width and height of the sensor. For the 50 Megapixel sensor with a pixel size of 4 µm (micrometer), the sensor has a width of 8820 pix * 4 µm/pix = 35,280 mm and a height of 5668 pix * 4 µm/pix = 22,672 mm.

The focal length f, the flight height h, the pixel size ps, and the amount of pixels pix determine the size of the area captured on the ground, so-called footprint, or swath s.

For a flight height of 1750 m, a focal length of 70 mm and the given sensor, the swath width is calculated by

sw = pix-hor * h * ps / f = (8820 pix * 1750 m * 4 µm/pix) / 70 mm = 882 m

and the swath height:

sh = pix-ver * h * ps / f = (5668 pix * 1750 m * 4 µm/pix) / 70 mm =566,8 m

The flight height h, the pixel size ps and the focal length f  also determine the area on the ground which is represented by one pixel. The distance between two consecutive pixels is called Ground Sampling Distance (GSD) and is calculated by

GSD = h* ps / f = (1750 m * 4 µm/pix) / 70 mm = 10 cm/pix

Images_Konrad.pdf

In most cases of photogrammetric projects the GSD is given and determines the flight height pending on the focal length of the used camera.

If we compare results that were taken at the same height, we will see that a higher resolution of the images can lead to more precise results, as the measurement accuracy is depending on the size of a pixel on the object (GSD).

  A large resolution sensor can also be beneficial in Photogrammetry as more feature points can be used to retrieve a stable camera orientation from Bundle Adjustment.

Most cameras use a Bayer pattern and effectively have a lower resolution than the nominal image. There is actually only half the pixels green, a quarter red and a quarter blue. The final image is retrieved by interpolation in the de-Bayering process.

Radiometric quality

The radiometric quality is the ability to resolve an intensity or brightness of a pixel and with that the sensitivity of the sensor. This quality is defined by the size of each pixel (e.g. 4.5 micrometer) as it can collect more light as well as the sensor quality. As Photogrammetry requires texture, the radiometric quality (e.g. the ability to resolve texture in low light conditions or shadows) is important for the final result. The final radiometric quality of the image is defined of the sensor quality and the amount of light hitting the pixel during exposure.

Sharpness

The sharpness is defined by the quality of the lens as well as the aperture being used. With a closed aperture (large value such as 11) the sharpness increases. The lens quality itself is defined by the manufacturing quality, the size of the lens (larger sensors can use larger lenses), the lens assembly (zoom lenses have typically less sharpness than fixed focal length lenses). Besides the lens itself, sharpness can also be diminished by motion blur, which should be considered with moving scenes by a short exposure time.


Exposure parameters

Choosing the right exposure parameters is crucial for image quality. For Photogrammetric applications, the image should ne neither over- or underexposed - so the histogram should be in the center. The following sequence of defining the parameters can be used: 1. Reduce the ISO value to a minimum (e.g. 100). The ISO value is an amplification of the sensor and leads on increase to more radiometric noise. 2. Close the aperture to a value of 5.6 to 8 depending on the light situation. High values lead to sharper images, higher values than 11 can cause aberration effects on some lenses. 3. Choose the exposure time of 1/100s or slightly shorter to retrieve a histogram in the center. If you have an application with motion involved, the exposure time should be chosen short enough to avoid motion blur. 4. In case of an underexposure adjust the aperture accordingly. (For more details on image acquisition find also this paper here, second last page)



Image Effects

Chromatic Aberration

Aberration is an effect that can be seen as little red or blue fringes at the corners of white edges - particularly in the corners of the image. It is caused by different inclinations that light is carried on through the lens based on its wavelength (color). As the wavelength is different of red, green and blue pixels, the same object is not placed at the exactly same place. The effect is radial and can be compensated by calibration automatically. If no correction is possible, the green channel only should be used in Photogrammetry as its wavelength is in the middle and typically has the best radiometric quality with Bayer-pattern cameras.

De-Bayering artifacts

De-Bayering artifacts occur as the interpolation process tries to retrieve a full sensor resolution image from the original capture. As there are always two pixels green, one blue and one red on the sensor, the final image is interpolated from composing them together. In order to preserve features such as edges, smart edge detections are used in some De-Bayering methods. In the presence of shadows or low radiometric signal-to-noise this can however lead to an increase of noise.

Vignetting

Vignetting leads to darker outer image regions. It is a circular effect caused by the lens directing less light to the sensors in these areas. It occurs particularly for lower quality lenses and can be removed by calibration.