求教一个 sql 优化问题

使用了 nodejs 的 orm 库 sequelize ，在一条查询中生成了如下 sql ，

SELECT `Images`.*, `tags`.`id` AS `tags.id`, `tags`.`name` AS `tags.name`, `tags`.`translatedName` AS `tags.translatedName`, `tags`.`customName` AS `tags.customName`, `author`.`id` AS `author.id`, `author`.`name` AS `author.name`, `author`.`profileImageUrl` AS `author.profileImageUrl`   


    FROM (SELECT `Images`.`id`, `Images`.`title`, `Images`.`previewUrl`, `Images`.`totalBookmarks`, `Images`.`totalView`, `Images`.`originUrlJson`, `Images`.`authorId` FROM `imgStorage` AS `Images`
            WHERE `Images`.`authorId` = 1096811 
            ORDER BY `Images`.`totalBookmarks` 
            DESC LIMIT '0', '30') AS `Images` 
    LEFT OUTER JOIN ( `ImagesTags` AS `tags->ImagesTags` INNER JOIN `imgTags` AS `tags` ON `tags`.`id` = `tags->ImagesTags`.`TagId`) 
    ON `Images`.`id` = `tags->ImagesTags`.`ImageId` 
    LEFT OUTER JOIN `Author` AS `author` ON `Images`.`authorId` = `author`.`id` ORDER BY `Images`.`totalBookmarks` DESC

EXPLAIN QUERY PLAN SELECT 发现会触发 ImagesTags AS tags->ImagesTags 的全表扫描，随着数据量增加会变的非常慢
几乎所有涉及 通过 Images ImagesTags 关联 Tag 的情况下都发生
于是尝试在 ImagesTags 建立如下索引 [ImageId ，TagId],[TagId,ImageId],[ImageId],[TagId]
发现不管是都建还是分别只建一个都不生效
于是想问下，这种情况有办法通过建立索引来解决么？如果可以应该怎么建？
我现在通过，单独查出 Images 然后每一项单独通过 ImagesTags 查 tags 能极大的解决性能问题，但后面如果想再通过 tag 来过滤 Images 似乎会很麻烦..求解