Academia.eduAcademia.edu

Figure 1 – uploaded by Stefano Rovetta

See full PDFdownloadDownload figure
‘ig. 1. Global schema of an automated license-plate recognition system

Figure 1 ‘ig. 1. Global schema of an automated license-plate recognition system

Related Figures (12)

Fig. 2. Basic VQ-based schema for image coding.
Fig. 2. Basic VQ-based schema for image coding.
Fig. 3. Size-adaptive schema of VQ-based image compression.
Fig. 3. Size-adaptive schema of VQ-based image compression.
Fig. 5. Global schema of the image-compression and license-plate location systems. The only parameter in such a process is the expected stripe size. The fact that the stripe width and height do not exhibit high variance values seems reasonable, as the eventual stripe size mostly depends on the relative sensor positioning (e.g., dis- tance from the scene), which is assumed to be known. This im- plies that training a stripe extractor just consists in averaging the stripe size from examples containing license plates.
Fig. 5. Global schema of the image-compression and license-plate location systems. The only parameter in such a process is the expected stripe size. The fact that the stripe width and height do not exhibit high variance values seems reasonable, as the eventual stripe size mostly depends on the relative sensor positioning (e.g., dis- tance from the scene), which is assumed to be known. This im- plies that training a stripe extractor just consists in averaging the stripe size from examples containing license plates.
Fig. 7. The VQ-encoding chip. (a) Internal layout. (b) Chip bonding.
Fig. 7. The VQ-encoding chip. (a) Internal layout. (b) Chip bonding.
Fig. 8. Comparison of VQ and JPEG compression qualities. The VQ schema performs better at higher compression ratios (typically, Cr > 35).
Fig. 8. Comparison of VQ and JPEG compression qualities. The VQ schema performs better at higher compression ratios (typically, Cr > 35).
Fig. 9. Visual nay tal of VQ and JPEG compression for picture archival purposes (Cr = 50). (a) Original picture. (b) VQ-compressed picture. (c) JPEG-compressed picture.
Fig. 9. Visual nay tal of VQ and JPEG compression for picture archival purposes (Cr = 50). (a) Original picture. (b) VQ-compressed picture. (c) JPEG-compressed picture.
Fig. 11. Result of adaptive block-size decomposition. Fig. 10. Input picture (digitized PAL field, 768 x 256, 8 bpp).
Fig. 11. Result of adaptive block-size decomposition. Fig. 10. Input picture (digitized PAL field, 768 x 256, 8 bpp).
Fig. 12. Compressed picture for archival storage.
Fig. 12. Compressed picture for archival storage.
Fig. 13. Candidate stripes extracted from Fig. 11 and their associate aspect scores.
Fig. 13. Candidate stripes extracted from Fig. 11 and their associate aspect scores.
fifa 644 STATISTICAL SUMMARY OF LOCATION RESULTS Results are summarized in Table II; they indicate that the overall system performance is quite satisfactory, as it exhibits a 2% error rate. This performance meets industrial requirements, as the specific OCR process adopted can detect the absence of text in a stripe and then trigger an exceptional type of pro- cessing. In practice, such critical pictures are stored in a special file for delayed visual inspection.
fifa 644 STATISTICAL SUMMARY OF LOCATION RESULTS Results are summarized in Table II; they indicate that the overall system performance is quite satisfactory, as it exhibits a 2% error rate. This performance meets industrial requirements, as the specific OCR process adopted can detect the absence of text in a stripe and then trigger an exceptional type of pro- cessing. In practice, such critical pictures are stored in a special file for delayed visual inspection.
Fig. 14. Visual representation of the image-region mapping. (a) Spatial-scoring map: darker areas correspond to smaller blocks. (b) VQ-scoring map: darker areas mark higher-score regions.
Fig. 14. Visual representation of the image-region mapping. (a) Spatial-scoring map: darker areas correspond to smaller blocks. (b) VQ-scoring map: darker areas mark higher-score regions.
Related topics:
Computer VisionArtificial VisionLicense-plate locationDocument Image AnalysisVector QuantizationLicense Plate RecognitionVector Quantisers
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved