Improving the Accuracy of Inexpensive Sensors for Optical Density Measurement

A common problem in the field of color printing is closed loop color calibration, a process which generally requires a spectrometer. Light Emitting Diodes (LED) sensors are inexpensive and effective for measuring the net reflectance of certain frequencies of light. Sensor measurement of LED reflectance can be used to estimate spectral reflectance or color value (measured in CIELAB or CIEXYZ color space) of a printed patch. LED based sensors suffer from their lower accuracy in predicting color values. This paper focuses on improving accuracy of these sensors in predicting Lightness values in CIELAB color space of a single colorant by combining information from multiple LEDs with different spectrum coverage. Introduction The lightness linearity of RGB space is critical in getting good image quality from multiple ink printers. Therefore, a common problem in the field of color printing is closed loop color calibration, a process which generally requires a spectrometer. Light Emitting Diodes (LED) sensors are inexpensive and effective for measuring the net reflectance of certain frequencies of light. Sensor measurement of LED reflectance can be used to estimate spectral reflectance or color value (measured in CIELAB or CIEXYZ color space) of a printed patch. However, these sensors suffer from several drawbacks: 1. Gain Control: Sensors do not receive enough light back for the color patches with high density and therefore they have a lower signal to noise ratio for dark patches. On the other hand, if the overall intensity of the LEDs is increased, the sensor may be saturated on light patches. 2. LED spectrum range is not optimized for certain types of inks, meaning that each LED may return only a portion of information about an ink density. 3. Different densities of the same ink do not necessarily have the same hue angle. This paper starts with an introduction to a general method for using LED based sensor to measure lightness of a printed patch in CIELAB color space. A new method is introduced to improve accuracy of these sensors. This method combines prediction of each LED based on the ink density and its reflectance match with the spectrum coverage of the LED. The accuracy of this method in predicting lightness of each color patch is compared to the conventional approaches in CIELAB color space. The result shows that the accuracy of the sensors in Color Calibration improves by as much as 50% in CIELAB color space. Close Loop Color Calibration Closed Loop Calibration in printers is used to get perpetually uniform steps of ink ramps. Most of the known calibration procedures ([1], [2]) adjust the density of individual inks through a lookup table. The main assumption is that the hue angle for different densities of an ink is constant. This means a correction based on lightness (L* in CIELAB) can give us perceptually linear ramps in RGB space. LED Based Sensors Performance of the calibration process in correcting density variation amongst pens in the printer is strongly dependent on the accuracy of measurement devices. Due to cost and smaller size, LED based sensors are common to be used to measure lightness (L*). In general, these sensors have 2 or more LEDs with different spectral range and one mono-chrome sensor that measures diffuse reflection of the light from the media. In our study we sued a sensor that is based on 4 LEDs (Red, Green, Blue and Orange) and has 2 sensors to detect diffuse and specular reflectance. Figure 1 shows position of diffuse and specular sensors respect to LED and surface. Figure 1: Sensor Design Measurement of diffuse channel is used to measure color characteristic of a surface. In close loop calibration, reading from this channel is used to predict L*. In the remaining sections, some existing models for predicting L* are compared. At the end a new method is introduced to improve the prediction performance. Using LED Based Sensor to Predict Lightness Each LED in the sensors covers specific wavelength range. Figure 2 shows the spectral range of each 4 LED for the sensor used in this paper. 0 0.2 0.4 0.6 0.8 1 1.2 38 0 40 0 42 0 44 0 46 0 48 0 50 0 52 0 54 0 56 0 58 0 60 0 62 0 64 0 66 0 68 0 70 0 72 0