Probl\u00e8me<\/th>	Correction<\/th><\/tr><\/thead>
absorption du rouge<\/td>	Red Channel Compensation<\/td><\/tr>
dominante bleu-vert<\/td>	Gray World<\/td><\/tr>
faible contraste<\/td>	CLAHE<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n R\u00e9sultat :<\/p>\n\n\n\n \n couleurs plus naturelles<\/li>\n\n\n\n textures des coraux plus visibles<\/li>\n\n\n\n segmentation automatique plus stable<\/li>\n<\/ul>\n\n\n\n Ce qui est tr\u00e8s utile pour les outils d\u2019analyse de r\u00e9cifs<\/strong><\/p>\n\n\n\n \n\n\n\n <\/span>Pourquoi ce script marche bien pour CoralReef<\/span><\/h2>\n\n\n\n Les outils d\u2019analyse de r\u00e9cifs utilisent souvent :<\/p>\n\n\n\n \n segmentation couleur<\/li>\n\n\n\n classification des coraux<\/li>\n<\/ul>\n\n\n\n Le script am\u00e9liore :<\/p>\n\n\n\n \n la s\u00e9paration corail \/ eau<\/strong><\/li>\n\n\n\n la visibilit\u00e9 des textures<\/strong><\/li>\n\n\n\n la coh\u00e9rence colorim\u00e9trique<\/strong><\/li>\n<\/ul>\n\n\n\n ce qui peut am\u00e9liorer la d\u00e9tection automatique<\/strong>.<\/p>\n\n\n\n \n\n\n\n NOTA: Si les images sont faites \u00e0 tr\u00e8s faible profondeur (~1m), vous pouvez r\u00e9duire l\u00e9g\u00e8rement la compensation rouge :<\/p>\n\n\n <\/div> <\/div> <\/div><\/div> <\/p>\n\n\n\n r = r + (g_mean - np.mean(r))*0.4\n<\/code><\/pre>\n\n\n<\/div><\/div>\n\n\n\n \n\n\n\nCe script Python peut \u00eatre utilis\u00e9 comme \u00e9tape de pr\u00e9-traitement des images pour l\u2019analyse automatique des r\u00e9cifs coralliens<\/strong>, par exemple avant l\u2019utilisation d\u2019outils de classification ou de segmentation.<\/p>\n\n\n\n \n","protected":false},"excerpt":{"rendered":"Sous l\u2019eau, les diff\u00e9rentes longueurs d\u2019onde sont progressivement absorb\u00e9es : le rouge dispara\u00eet en premier, puis l\u2019orange et le jaune. M\u00eame \u00e0 faible profondeur, les photos pr\u00e9sentent souvent une dominante bleu-cyan. R\u00e9sultat : les photos paraissent…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"give_campaign_id":0,"_bbp_topic_count":0,"_bbp_reply_count":0,"_bbp_total_topic_count":0,"_bbp_total_reply_count":0,"_bbp_voice_count":0,"_bbp_anonymous_reply_count":0,"_bbp_topic_count_hidden":0,"_bbp_reply_count_hidden":0,"_bbp_forum_subforum_count":0,"sfsi_plus_gutenberg_text_before_share":"","sfsi_plus_gutenberg_show_text_before_share":"","sfsi_plus_gutenberg_icon_type":"","sfsi_plus_gutenberg_icon_alignemt":"","sfsi_plus_gutenburg_max_per_row":"","_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[147,222],"tags":[4446,4484,4436,4444,4438,131,4437],"class_list":["post-16489","post","type-post","status-publish","format-standard","hentry","category-images","category-python","tag-computer-vision","tag-coral-reefs-fr","tag-image-processing","tag-marine-science","tag-opencv","tag-python","tag-underwater-photography"],"aioseo_notices":[],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p6XU0A-4hX","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts\/16489","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/comments?post=16489"}],"version-history":[{"count":0,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts\/16489\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/media?parent=16489"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/categories?post=16489"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/tags?post=16489"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}

Contenu <\/p>\nToggle<\/span>

Script Python<\/a><\/li>
Installation<\/a><\/li>
Structure des dossiers<\/a><\/li>
Pourquoi cette m\u00e9thode est utilis\u00e9e en recherche<\/a><\/li>
Pourquoi ce script marche bien pour CoralReef<\/a><\/li><\/ul><\/nav><\/div>\n\n\n\n\n
\n\n\n\n
\u00c0 faible profondeur<\/strong> (<2-3m), la perte de rouge est faible mais l\u2019image est souvent domin\u00e9e par le bleu\/cyan<\/strong>.
Pour un traitement automatique (avant CoralReef<\/strong>, par exemple), la meilleure approche est :<\/p>\n\n\n\n
1. augmentation du contraste<\/strong><\/li>\n<\/ol>\n\n\n\n
  On peut faire cela facilement avec Python + OpenCV<\/strong> sur tout un r\u00e9pertoire.<\/p>\n\n\n\n
  En traitement scientifique d\u2019images sous-marines, plusieurs m\u00e9thodes issues de la vision par ordinateur sont couramment utilis\u00e9es :<\/p>\n\n\n\n
  - parfois Retinex<\/strong> pour corriger l\u2019illumination<\/li>\n<\/ul>\n\n\n\n
    Cette combinaison est utilis\u00e9e dans plusieurs travaux de recherche sur les r\u00e9cifs coralliens et les robots sous-marins<\/strong>.<\/p>\n\n\n\n
    Je vous propose ici un script Python robuste qui applique :<\/p>\n\n\n\n
    1️⃣ compensation du canal rouge<\/strong>
    2️⃣ balance des blancs Gray-World<\/strong>
    3️⃣ am\u00e9lioration du contraste (CLAHE)<\/strong>
    4️⃣ traitement batch d\u2019un dossier<\/strong><\/p>\n\n\n\n
    C\u2019est nettement plus performant<\/strong> que les scripts simples.<\/p>\n\n\n\n
    \n\n\n\n<\/span>Script Python <\/span><\/h2>\n\n\n\n
    Ce script :<\/p>\n\n\n\n
    \n
    lit toutes les images d\u2019un dossier<\/li>\n\n\n\n
    corrige la balance des blancs<\/li>\n\n\n\n
    r\u00e9\u00e9quilibre les canaux couleur<\/li>\n\n\n\n
    enregistre les images corrig\u00e9es dans un nouveau dossier<\/li>\n<\/ul>\n\n\n
    <\/div>
    <\/div>
    <\/div><\/div>
    <\/p>\n\n\n\n
    import cv2\nimport numpy as np\nimport os\nfrom pathlib import Path\n\ninput_dir = \"photos_brutes\"\noutput_dir = \"photos_corrigees\"\n\nPath(output_dir).mkdir(exist_ok=True)\n\n# -----------------------------\n# Red Channel Compensation\n# -----------------------------\ndef red_channel_compensation(img):\n\n b,g,r = cv2.split(img)\n\n b_mean = np.mean(b)\n g_mean = np.mean(g)\n\n r = r + (g_mean - np.mean(r))*0.6\n r = r + (b_mean - np.mean(r))*0.3\n\n r = np.clip(r,0,255)\n\n return cv2.merge([b,g,r.astype(np.uint8)])\n\n\n# -----------------------------\n# Gray World White Balance\n# -----------------------------\ndef gray_world(img):\n\n img = img.astype(np.float32)\n\n avg_b = np.mean(img[:,:,0])\n avg_g = np.mean(img[:,:,1])\n avg_r = np.mean(img[:,:,2])\n\n avg = (avg_b + avg_g + avg_r) \/ 3\n\n img[:,:,0] *= avg\/avg_b\n img[:,:,1] *= avg\/avg_g\n img[:,:,2] *= avg\/avg_r\n\n img = np.clip(img,0,255)\n\n return img.astype(np.uint8)\n\n\n# -----------------------------\n# CLAHE Contrast Enhancement\n# -----------------------------\ndef enhance_contrast(img):\n\n lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)\n\n l,a,b = cv2.split(lab)\n\n clahe = cv2.createCLAHE(clipLimit=2.5, tileGridSize=(8,8))\n l = clahe.apply(l)\n\n lab = cv2.merge((l,a,b))\n\n return cv2.cvtColor(lab, cv2.COLOR_LAB2BGR)\n\n\n# -----------------------------\n# Pipeline complet\n# -----------------------------\ndef process(img):\n\n img = red_channel_compensation(img)\n img = gray_world(img)\n img = enhance_contrast(img)\n\n return img\n\n\n# -----------------------------\n# Batch processing\n# -----------------------------\nfor file in os.listdir(input_dir):\n\n if file.lower().endswith((\".jpg\",\".jpeg\",\".png\",\".tif\",\".tiff\")):\n\n path = os.path.join(input_dir,file)\n\n img = cv2.imread(path)\n\n if img is None:\n print(\"image ignor\u00e9e:\", file)\n continue\n\n corrected = process(img)\n\n outpath = os.path.join(output_dir,file)\n\n cv2.imwrite(outpath,corrected)\n\n print(\"processed:\",file)\n\nprint(\"Termin\u00e9\")<\/code><\/pre>\n\n\n<\/div><\/div>\n\n\n\n \n\n\n\n<\/span>Installation<\/span><\/h2>\n\n\n\nDans un fen\u00eatre shell OsGeo4W:<\/p>\n\n\n <\/div><\/div><\/div><\/div><\/p>\n\n\n\npip install opencv-python numpy\n<\/code><\/pre>\n\n\n<\/div><\/div>\n\n\n\n \n\n\n\n<\/span>Structure des dossiers<\/span><\/h2>\n\n\n\nprojet\/\n script.py\n photos_brutes\/\n img1.jpg\n img2.jpg\n photos_corrigees\/<\/code><\/pre>\n\n\n\n \n\n\n\n<\/a>Avant correction avec le script Apr\u00e8s correction<\/figcaption><\/figure>\n<\/div>\n\n\n \n\n\n\n<\/span>Pourquoi cette m\u00e9thode est utilis\u00e9e en recherche<\/span><\/h2>\n\n\n\nElle corrige les trois probl\u00e8mes majeurs des images sous-marines<\/strong> :<\/p>\n\n\n\n Probl\u00e8me<\/th> Correction<\/th><\/tr><\/thead> absorption du rouge<\/td> Red Channel Compensation<\/td><\/tr> dominante bleu-vert<\/td> Gray World<\/td><\/tr> faible contraste<\/td> CLAHE<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\nR\u00e9sultat :<\/p>\n\n\n\n \ncouleurs plus naturelles<\/li>\n\n\n\n textures des coraux plus visibles<\/li>\n\n\n\n segmentation automatique plus stable<\/li>\n<\/ul>\n\n\n\nCe qui est tr\u00e8s utile pour les outils d\u2019analyse de r\u00e9cifs<\/strong><\/p>\n\n\n\n \n\n\n\n<\/span>Pourquoi ce script marche bien pour CoralReef<\/span><\/h2>\n\n\n\nLes outils d\u2019analyse de r\u00e9cifs utilisent souvent :<\/p>\n\n\n\n \nsegmentation couleur<\/li>\n\n\n\n classification des coraux<\/li>\n<\/ul>\n\n\n\nLe script am\u00e9liore :<\/p>\n\n\n\n \nla s\u00e9paration corail \/ eau<\/strong><\/li>\n\n\n\n la visibilit\u00e9 des textures<\/strong><\/li>\n\n\n\n la coh\u00e9rence colorim\u00e9trique<\/strong><\/li>\n<\/ul>\n\n\n\nce qui peut am\u00e9liorer la d\u00e9tection automatique<\/strong>.<\/p>\n\n\n\n \n\n\n\nNOTA: Si les images sont faites \u00e0 tr\u00e8s faible profondeur (~1m), vous pouvez r\u00e9duire l\u00e9g\u00e8rement la compensation rouge :<\/p>\n\n\n <\/div><\/div><\/div><\/div><\/p>\n\n\n\nr = r + (g_mean - np.mean(r))*0.4\n<\/code><\/pre>\n\n\n<\/div><\/div>\n\n\n\n \n\n\n\nCe script Python peut \u00eatre utilis\u00e9 comme \u00e9tape de pr\u00e9-traitement des images pour l\u2019analyse automatique des r\u00e9cifs coralliens<\/strong>, par exemple avant l\u2019utilisation d\u2019outils de classification ou de segmentation.<\/p>\n\n\n\n \n","protected":false},"excerpt":{"rendered":"Sous l\u2019eau, les diff\u00e9rentes longueurs d\u2019onde sont progressivement absorb\u00e9es : le rouge dispara\u00eet en premier, puis l\u2019orange et le jaune. M\u00eame \u00e0 faible profondeur, les photos pr\u00e9sentent souvent une dominante bleu-cyan. R\u00e9sultat : les photos paraissent…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"give_campaign_id":0,"_bbp_topic_count":0,"_bbp_reply_count":0,"_bbp_total_topic_count":0,"_bbp_total_reply_count":0,"_bbp_voice_count":0,"_bbp_anonymous_reply_count":0,"_bbp_topic_count_hidden":0,"_bbp_reply_count_hidden":0,"_bbp_forum_subforum_count":0,"sfsi_plus_gutenberg_text_before_share":"","sfsi_plus_gutenberg_show_text_before_share":"","sfsi_plus_gutenberg_icon_type":"","sfsi_plus_gutenberg_icon_alignemt":"","sfsi_plus_gutenburg_max_per_row":"","_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[147,222],"tags":[4446,4484,4436,4444,4438,131,4437],"class_list":["post-16489","post","type-post","status-publish","format-standard","hentry","category-images","category-python","tag-computer-vision","tag-coral-reefs-fr","tag-image-processing","tag-marine-science","tag-opencv","tag-python","tag-underwater-photography"],"aioseo_notices":[],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p6XU0A-4hX","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts\/16489","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/comments?post=16489"}],"version-history":[{"count":0,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts\/16489\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/media?parent=16489"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/categories?post=16489"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/tags?post=16489"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}

<\/span>Script Python <\/span><\/h2>\n\n\n\nCe script :<\/p>\n\n\n\n\nlit toutes les images d\u2019un dossier<\/li>\n\n\n\ncorrige la balance des blancs<\/li>\n\n\n\nr\u00e9\u00e9quilibre les canaux couleur<\/li>\n\n\n\n