ISET simulations for another cmos imager

From VISTA LAB WIKI

Jump to: navigation, search

These are the simulation parameters that we will use to evaluate the effect that pixel size, scene luminance and read noise and subjective preferences.

SCENE: The scene data will be radiance images generated using the LED multispectral imaging system developed by Max Klein, Manu Parmar and Joyce Farrell.

OPTICS: The optics data will be determined by a diffraction-limited shift-invariant model. In one condition, the lens f-number will be 2.0 and the focal length will be 3.7. There will be a separate study in which we optimize the lens f-number given prior work by Peter Catrysse and his students.

SENSOR: The sensor data are based on data obtained from Aptina for a sensor with a 2.8 micron pixel. We use the Aptina data on well capacity and conversion gain to determine the voltage swing. (conversion gain * well capacity = voltage swing).

We use the Aptina data for dark voltage and read noise. Aptina did not provide data on prnu and dsnu, so we estimated these values from our own measurements.

The sensor properties for the other sensors (with pixel size of 2.2, 1.75, 1.4 and 1.2) are based on the following assumptions. 1. Voltage swing is the same (1.28) 2. Well capacity is proportional to pixel area. We use the well capacity for the 2.8 sensor (17500) and the pixel size to calculate the proportionality 17500/(2.8^2) = 2232.1/micron^2 (2.2^2 * 2232.1 = 10804 electrons for the 2.2 (1.75^2* 2232.1 = 6836 electrons for the 1.75 (1.4^2* 2232.1 = 4375 electrons for the 1.4 (1.2^2* 2232.1 = 3214 electrons for the 1.2 3. conversion gain = voltage swing / well capacity 4. Fill factor is 90% - assuming that the microlens is effective in focusing light on the active region of each pixel

We treat read noise as a parameter. For each sensor, we will consider read noise values of 2.5, 5, 10 , 20, 40 electron. We use the conversion gain to convert read noise in electrons to read noise in mV.

Parameters

 

 

 

 

 

Pixel size (microns)

2.8

2.2

1.75

1.4

1.2

F number1

2.0

2.0

2.0

2.0

2.0

Focal length1

3.7

3.7

3.7

3.7

3.7

Well capacity(e)5

17500

10804

6836

4375

3214

Conversion gain (uV/e)7

73

118

187

293

398

Voltage swing (V)6

1.28

1.28

1.28

1.28

1.28

Fill factor3

90%

90%

90%

90%

90%

Dark voltage (mV/pixel/sec)4

4.38

4.38

4.38

4.38

4.38

Read noise (mV) 4 

See below

See below

See below

See below

See below

DSNU (mV) 4

0.83

0.83

0.83

0.83

0.83

PRNU (%)4

0.91

0.91

0.91

0.91

0.91

 

 

 

 

 

 

Read noise in mV for 2.5 electrons

0.1823

0.2950

  0.4675

0.7325

0.995

Read noise in mV for
5 electrons

0.3650

0.5900

0.9350

1.465

1.99

Read noise in mV for
10 electrons

0.73

1.1800

1.870

2.93

3.98

Read noise in mV for
20 electrons

1.46

2.3600

3.74

5.86

7.96

Read noise in mV for
40 electrons

2.92

4.7200

7.48

11.72

15.92


PROCESSOR

Color correction: We use radiance scene data for a Macbeth ColorChecker (MCC) as a scene captured by each sensor. We then calculate the optimal color correction matrix (CCM) based on the sensor image of the MCC. The method we used is described http://www.imageval.com/public/Products/ISET/ApplicationNotes/ColorCorrectionMatrix.pdf

Demosaicing: We use bilinear demosaicing.

Downsampling: The demosaicked sensor images are downsampled so that they can be viewed as the same size images (NxM) on a display. The method we use to downsample ….

Personal tools