Ëàáîðàòîðèÿ êîìïüþòåðíîé ãðàôèêè è ìóëüòèìåäèà

Introduction

This project is dedicated to enhancement of microscopic images and related problems in order to simplify further analysis (by human or by computer) of such images. The system of regular optical microscope with programmatically controllable camera plugged on it is considered. Image from camera is processed on computer and is finally displayed on operator’s monitor. Microscope image analysis is wide-spreaded in many areas including biology, medicine, manufacturing quality control etc. However, images acquired using described system, have a number of drawbacks, such as limited dynamic range of luminance or limited depth of field. This project is aimed on development of hardware-software tools allowing to remove these drawbacks.

Drawbacks of microscopic images

The main drawbacks of described system are:

1. Dynamic range of conventional digital cameras and monitors is quite small. One may assume it to be 400:1 for cameras and 100:1 for monitors. Meanwhile dynamic range of human eye is assumed to be 10⁹:1 (considering accommodation) and 10000:1 (in frames of single scene). Hence an image on the screen appears much less contrast then an image that is seen directly through the eyepieces of the microscope.
2. Because of physical limitations of optics even the most expensive microscope has very small depth of field under significant magnification. It means that only the small part of observed object would be focused on image. Those parts of object placed out of the depth of field would appear blurry. And the further the area is placed from the depth of field the more blurry it appears (see pic. 1).
3. Using the stereo-microscope user is able to recognize 3D-structure of the object. But there usually is only one camera and no such information is available while processing.

Proposed methods

In our “Digital Microscope” project we propose the following methods to get rid of mentioned drawbacks:

1. Contrast enhancement with HDRI technology (High Dynamic Range Imaging). A number of images taken with different exposures is done and the image with high dynamic range is built. Then the tone mapping of HDR-image is done preparing it for output on conventional monitor while preserving visual contrast (see pic. 2).
2. Virtual depth of field extension with help of EDF algorithms (Extended Depth of Field). Microscope is placed on motorized, programmatically controlled shove allowing to move it up and down on specified distance. The number of images taken on different positions of camera is done. Different parts of object appear in focus on different images. The sharpest parts are selected from different images and combined together on resulting image with extended depth of field (see pic. 3).
3. Building of 3D-model using SFF methods (Shape From Focus) and stereo-monitor output. Based on information about sharpness measure of particular regions on different images, depth map of the object is built (SFF method). Depth map is then used to build 3D-model on which the EDF-image is put. Acquired model is displayed on stereo-monitor and user is allowed to observe 3D picture and even rotate the model without any actions performed with real object (see pic. 4).

Pic. 1: Example of the source image acquired directly from camera with automatically selected exposure	Pic. 2: Image of the same scene, enhanced with help of HDRI technology

Pic. 3: Image of the same scene with virtually extended depth of field be usage of EDF methods	Pic. 4: 3D-model of the scene, built using SFF methods

Ó÷àñòíèêè

• Ìèõàèë Ìàòðîñîâ, Lomonosov Moscow State University
• Âèêòîð Ãàãàíîâ, Lomonosov Moscow State University
• Àëåêñåé Âèêòîðîâè÷ Èãíàòåíêî, ÌÃÓ èì. Ëîìîíîñîâà

Ïóáëèêàöèè

• Ìàòðîñîâ , Ãàãàíîâ , Èãíàòåíêî , Ñèâîâîëåíêî . Óëó÷øåíèå êà÷åñòâà èçîáðàæåíèÿ ñ ìèêðîñêîïà ïðè ïîìîùè òåõíîëîãèè HDRI â èíòåðàêòèâíîì ðåæèìå. GraphiCon'2009, 2009 (download)