We present a novel method for fusing the results of multiple semantic video indexing algorithms that use different
types of feature descriptors and different classification methods. This method, called Context-Dependent Fusion
(CDF), is motivated by the fact that the relative performance of different semantic indexing methods can vary
significantly depending on the video type, context information, and the high-level concept of the video segment
to be labeled. The training part of CDF has two main components: context extraction and algorithm fusion.
In context extraction, the low-level audio-visual descriptors used by the different classification algorithms are
combined and used to partition the descriptors space into groups of similar video shots, or contexts. The
algorithm fusion component identifies a subset of classification algorithms (local experts) for each context based
on their relative performance within the context. Results on the TRECVID-2002 data collections show that the
proposed method can identify meaningful and coherent clusters and that different labeling algorithms can be
identified for the different contexts. Our initial experiments have indicated that the context-dependent fusion
outperforms the individual algorithms. We also show that using simple visual descriptors and a simple K-NN
classifier, the CDF approach provides results that are comparable to the state-of-the-art methods in semantic
The laser cleaning of the photoresist (PR) layer has been investigated as a function of laser energy density. The cleaning of the PR layer on silicon wafer was performed by a line beam of a KrF excimer laser in a cleanroom environment and then the applied energy density was 100 - 300 mJ/cm<sup>2</sup>. The experimental results showed that the ablation rates of the PR are increased with increasing of laser energy density without silicon wafer damage. The ablation rates of PR were 0.09 μm/pulse for 100 mJ/cm<sup>2</sup>, 0.15 μm/pulse for 200 mJ/cm<sup>2</sup> and 0.19 μm/pulse for 300 mJ/cm<sup>2</sup> with repetition rate of 30 Hz. The compositions of the PR covered wafers before and after laser irradiation were determined by Fourier transform infrared spectroscopy (FT-IR). The comparison of the cleaning results done in applying the laser cleaning to remove the PR and the metallic polymers resulting from reactive ion etching (RIE) was made before and after laser irradiation by scanning electron microscope (SEM). It is also shown that the PR and metallic polymer in the contact hole can be completely removed by the laser cleaning technique.
The importance of surface cleaning is an essential factor in VLSI technology, flat panel display, and data storage devices. The results of laser cleaning technology were studied using KrF excimer laser (248 nm) irradiation in cleanroom environment. The applied energy density was 200 - 800 mJ/cm<SUP>2</SUP> at a repetition rate of 10 - 40 Hz with various focused beam widths. Results of photoresist stripping were made before and after laser irradiation with PR covered wafers and comparison of laser cleaning results were investigated as well with bare wafers. The atomic force microscopy (AFM) images of laser cleaning results were also presented and compared before and after laser irradiation. The surface roughness of AFM image of contaminated wafer surface before laser irradiation was 192 angstrom and that of after laser irradiation was 16.2 angstrom. The mechanism of laser cleaning and ablation is rapid thermal expansion of substrate surface induced by an instantaneous temperature rising due to laser irradiation. It is found that the temperature rising of the substrate surface was about 297 degree(s)C with a fluence of 400 mJ/cm<SUP>2</SUP> at 300K. Laser dry cleaning technology easily removed fingerprints, submicron Al<SUB>2</SUB>O<SUB>3</SUB> and SiO<SUB>2</SUB> particulates intentionally contaminated on the top of the wafer surface without aids of toxic chemicals and deionized water.